diff options
Diffstat (limited to 'sys')
| -rw-r--r-- | sys/amd64/linux/linux.h | 7 | ||||
| -rw-r--r-- | sys/amd64/linux/linux_machdep.c | 169 | ||||
| -rw-r--r-- | sys/amd64/linux32/linux.h | 7 | ||||
| -rw-r--r-- | sys/amd64/linux32/linux32_machdep.c | 184 | ||||
| -rw-r--r-- | sys/compat/linux/linux_mmap.c | 257 | ||||
| -rw-r--r-- | sys/compat/linux/linux_mmap.h | 49 | ||||
| -rw-r--r-- | sys/i386/linux/linux.h | 7 | ||||
| -rw-r--r-- | sys/i386/linux/linux_machdep.c | 188 | ||||
| -rw-r--r-- | sys/modules/linux/Makefile | 2 | ||||
| -rw-r--r-- | sys/modules/linux_common/Makefile | 2 |
10 files changed, 318 insertions, 554 deletions
diff --git a/sys/amd64/linux/linux.h b/sys/amd64/linux/linux.h index 639499ab58d0..b30568ffaa3c 100644 --- a/sys/amd64/linux/linux.h +++ b/sys/amd64/linux/linux.h @@ -139,13 +139,6 @@ struct l_rlimit { l_ulong rlim_max; }; -/* mmap options */ -#define LINUX_MAP_SHARED 0x0001 -#define LINUX_MAP_PRIVATE 0x0002 -#define LINUX_MAP_FIXED 0x0010 -#define LINUX_MAP_ANON 0x0020 -#define LINUX_MAP_GROWSDOWN 0x0100 - /* * stat family of syscalls */ diff --git a/sys/amd64/linux/linux_machdep.c b/sys/amd64/linux/linux_machdep.c index 0459ba7afd15..80c48aa68944 100644 --- a/sys/amd64/linux/linux_machdep.c +++ b/sys/amd64/linux/linux_machdep.c @@ -83,6 +83,7 @@ __FBSDID("$FreeBSD$"); #include <compat/linux/linux_ipc.h> #include <compat/linux/linux_file.h> #include <compat/linux/linux_misc.h> +#include <compat/linux/linux_mmap.h> #include <compat/linux/linux_signal.h> #include <compat/linux/linux_util.h> #include <compat/linux/linux_emul.h> @@ -122,181 +123,19 @@ linux_set_upcall_kse(struct thread *td, register_t stack) return (0); } -#define STACK_SIZE (2 * 1024 * 1024) -#define GUARD_SIZE (4 * PAGE_SIZE) - int linux_mmap2(struct thread *td, struct linux_mmap2_args *args) { - struct proc *p = td->td_proc; - struct mmap_args /* { - caddr_t addr; - size_t len; - int prot; - int flags; - int fd; - long pad; - off_t pos; - } */ bsd_args; - int error; - struct file *fp; - cap_rights_t rights; - - LINUX_CTR6(mmap2, "0x%lx, %ld, %ld, 0x%08lx, %ld, 0x%lx", - args->addr, args->len, args->prot, - args->flags, args->fd, args->pgoff); - - error = 0; - bsd_args.flags = 0; - fp = NULL; - - /* - * Linux mmap(2): - * You must specify exactly one of MAP_SHARED and MAP_PRIVATE - */ - if (! ((args->flags & LINUX_MAP_SHARED) ^ - (args->flags & LINUX_MAP_PRIVATE))) - return (EINVAL); - - if (args->flags & LINUX_MAP_SHARED) - bsd_args.flags |= MAP_SHARED; - if (args->flags & LINUX_MAP_PRIVATE) - bsd_args.flags |= MAP_PRIVATE; - if (args->flags & LINUX_MAP_FIXED) - bsd_args.flags |= MAP_FIXED; - if (args->flags & LINUX_MAP_ANON) - bsd_args.flags |= MAP_ANON; - else - bsd_args.flags |= MAP_NOSYNC; - if (args->flags & LINUX_MAP_GROWSDOWN) - bsd_args.flags |= MAP_STACK; - /* - * PROT_READ, PROT_WRITE, or PROT_EXEC implies PROT_READ and PROT_EXEC - * on Linux/i386. We do this to ensure maximum compatibility. - * Linux/ia64 does the same in i386 emulation mode. - */ - bsd_args.prot = args->prot; - if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC)) - bsd_args.prot |= PROT_READ | PROT_EXEC; - - /* Linux does not check file descriptor when MAP_ANONYMOUS is set. */ - bsd_args.fd = (bsd_args.flags & MAP_ANON) ? -1 : args->fd; - if (bsd_args.fd != -1) { - /* - * Linux follows Solaris mmap(2) description: - * The file descriptor fildes is opened with - * read permission, regardless of the - * protection options specified. - */ - - error = fget(td, bsd_args.fd, - cap_rights_init(&rights, CAP_MMAP), &fp); - if (error != 0 ) - return (error); - if (fp->f_type != DTYPE_VNODE) { - fdrop(fp, td); - return (EINVAL); - } - - /* Linux mmap() just fails for O_WRONLY files */ - if (!(fp->f_flag & FREAD)) { - fdrop(fp, td); - return (EACCES); - } - - fdrop(fp, td); - } - - if (args->flags & LINUX_MAP_GROWSDOWN) { - /* - * The Linux MAP_GROWSDOWN option does not limit auto - * growth of the region. Linux mmap with this option - * takes as addr the initial BOS, and as len, the initial - * region size. It can then grow down from addr without - * limit. However, Linux threads has an implicit internal - * limit to stack size of STACK_SIZE. Its just not - * enforced explicitly in Linux. But, here we impose - * a limit of (STACK_SIZE - GUARD_SIZE) on the stack - * region, since we can do this with our mmap. - * - * Our mmap with MAP_STACK takes addr as the maximum - * downsize limit on BOS, and as len the max size of - * the region. It then maps the top SGROWSIZ bytes, - * and auto grows the region down, up to the limit - * in addr. - * - * If we don't use the MAP_STACK option, the effect - * of this code is to allocate a stack region of a - * fixed size of (STACK_SIZE - GUARD_SIZE). - */ - - if ((caddr_t)PTRIN(args->addr) + args->len > - p->p_vmspace->vm_maxsaddr) { - /* - * Some Linux apps will attempt to mmap - * thread stacks near the top of their - * address space. If their TOS is greater - * than vm_maxsaddr, vm_map_growstack() - * will confuse the thread stack with the - * process stack and deliver a SEGV if they - * attempt to grow the thread stack past their - * current stacksize rlimit. To avoid this, - * adjust vm_maxsaddr upwards to reflect - * the current stacksize rlimit rather - * than the maximum possible stacksize. - * It would be better to adjust the - * mmap'ed region, but some apps do not check - * mmap's return value. - */ - PROC_LOCK(p); - p->p_vmspace->vm_maxsaddr = (char *)USRSTACK - - lim_cur_proc(p, RLIMIT_STACK); - PROC_UNLOCK(p); - } - - /* - * This gives us our maximum stack size and a new BOS. - * If we're using VM_STACK, then mmap will just map - * the top SGROWSIZ bytes, and let the stack grow down - * to the limit at BOS. If we're not using VM_STACK - * we map the full stack, since we don't have a way - * to autogrow it. - */ - if (args->len > STACK_SIZE - GUARD_SIZE) { - bsd_args.addr = (caddr_t)PTRIN(args->addr); - bsd_args.len = args->len; - } else { - bsd_args.addr = (caddr_t)PTRIN(args->addr) - - (STACK_SIZE - GUARD_SIZE - args->len); - bsd_args.len = STACK_SIZE - GUARD_SIZE; - } - } else { - bsd_args.addr = (caddr_t)PTRIN(args->addr); - bsd_args.len = args->len; - } - bsd_args.pos = (off_t)args->pgoff; - - error = sys_mmap(td, &bsd_args); - - LINUX_CTR2(mmap2, "return: %d (%p)", - error, td->td_retval[0]); - return (error); + return (linux_mmap_common(td, PTROUT(args->addr), args->len, args->prot, + args->flags, args->fd, args->pgoff)); } int linux_mprotect(struct thread *td, struct linux_mprotect_args *uap) { - struct mprotect_args bsd_args; - - LINUX_CTR(mprotect); - bsd_args.addr = uap->addr; - bsd_args.len = uap->len; - bsd_args.prot = uap->prot; - if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC)) - bsd_args.prot |= PROT_READ | PROT_EXEC; - return (sys_mprotect(td, &bsd_args)); + return (linux_mprotect_common(td, PTROUT(uap->addr), uap->len, uap->prot)); } int diff --git a/sys/amd64/linux32/linux.h b/sys/amd64/linux32/linux.h index 02d12f554e8e..97da878339c0 100644 --- a/sys/amd64/linux32/linux.h +++ b/sys/amd64/linux32/linux.h @@ -165,13 +165,6 @@ struct l_rusage { l_long ru_nivcsw; } __packed; -/* mmap options */ -#define LINUX_MAP_SHARED 0x0001 -#define LINUX_MAP_PRIVATE 0x0002 -#define LINUX_MAP_FIXED 0x0010 -#define LINUX_MAP_ANON 0x0020 -#define LINUX_MAP_GROWSDOWN 0x0100 - struct l_mmap_argv { l_uintptr_t addr; l_size_t len; diff --git a/sys/amd64/linux32/linux32_machdep.c b/sys/amd64/linux32/linux32_machdep.c index 25f023c741aa..187ec15e0218 100644 --- a/sys/amd64/linux32/linux32_machdep.c +++ b/sys/amd64/linux32/linux32_machdep.c @@ -70,6 +70,7 @@ __FBSDID("$FreeBSD$"); #include <amd64/linux32/linux32_proto.h> #include <compat/linux/linux_ipc.h> #include <compat/linux/linux_misc.h> +#include <compat/linux/linux_mmap.h> #include <compat/linux/linux_signal.h> #include <compat/linux/linux_util.h> #include <compat/linux/linux_emul.h> @@ -84,9 +85,6 @@ struct l_old_select_argv { l_uintptr_t timeout; } __packed; -static int linux_mmap_common(struct thread *td, l_uintptr_t addr, - l_size_t len, l_int prot, l_int flags, l_int fd, - l_loff_t pos); static void bsd_to_linux_rusage(struct rusage *ru, struct l_rusage *lru) @@ -448,9 +446,6 @@ linux_set_upcall_kse(struct thread *td, register_t stack) return (0); } -#define STACK_SIZE (2 * 1024 * 1024) -#define GUARD_SIZE (4 * PAGE_SIZE) - int linux_mmap2(struct thread *td, struct linux_mmap2_args *args) { @@ -489,184 +484,11 @@ linux_mmap(struct thread *td, struct linux_mmap_args *args) (uint32_t)linux_args.pgoff)); } -static int -linux_mmap_common(struct thread *td, l_uintptr_t addr, l_size_t len, l_int prot, - l_int flags, l_int fd, l_loff_t pos) -{ - struct proc *p = td->td_proc; - struct mmap_args /* { - caddr_t addr; - size_t len; - int prot; - int flags; - int fd; - long pad; - off_t pos; - } */ bsd_args; - int error; - struct file *fp; - cap_rights_t rights; - - error = 0; - bsd_args.flags = 0; - fp = NULL; - - /* - * Linux mmap(2): - * You must specify exactly one of MAP_SHARED and MAP_PRIVATE - */ - if (!((flags & LINUX_MAP_SHARED) ^ (flags & LINUX_MAP_PRIVATE))) - return (EINVAL); - - if (flags & LINUX_MAP_SHARED) - bsd_args.flags |= MAP_SHARED; - if (flags & LINUX_MAP_PRIVATE) - bsd_args.flags |= MAP_PRIVATE; - if (flags & LINUX_MAP_FIXED) - bsd_args.flags |= MAP_FIXED; - if (flags & LINUX_MAP_ANON) { - /* Enforce pos to be on page boundary, then ignore. */ - if ((pos & PAGE_MASK) != 0) - return (EINVAL); - pos = 0; - bsd_args.flags |= MAP_ANON; - } else - bsd_args.flags |= MAP_NOSYNC; - if (flags & LINUX_MAP_GROWSDOWN) - bsd_args.flags |= MAP_STACK; - - /* - * PROT_READ, PROT_WRITE, or PROT_EXEC implies PROT_READ and PROT_EXEC - * on Linux/i386. We do this to ensure maximum compatibility. - * Linux/ia64 does the same in i386 emulation mode. - */ - bsd_args.prot = prot; - if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC)) - bsd_args.prot |= PROT_READ | PROT_EXEC; - - /* Linux does not check file descriptor when MAP_ANONYMOUS is set. */ - bsd_args.fd = (bsd_args.flags & MAP_ANON) ? -1 : fd; - if (bsd_args.fd != -1) { - /* - * Linux follows Solaris mmap(2) description: - * The file descriptor fildes is opened with - * read permission, regardless of the - * protection options specified. - */ - - error = fget(td, bsd_args.fd, - cap_rights_init(&rights, CAP_MMAP), &fp); - if (error != 0) - return (error); - if (fp->f_type != DTYPE_VNODE) { - fdrop(fp, td); - return (EINVAL); - } - - /* Linux mmap() just fails for O_WRONLY files */ - if (!(fp->f_flag & FREAD)) { - fdrop(fp, td); - return (EACCES); - } - - fdrop(fp, td); - } - - if (flags & LINUX_MAP_GROWSDOWN) { - /* - * The Linux MAP_GROWSDOWN option does not limit auto - * growth of the region. Linux mmap with this option - * takes as addr the initial BOS, and as len, the initial - * region size. It can then grow down from addr without - * limit. However, Linux threads has an implicit internal - * limit to stack size of STACK_SIZE. Its just not - * enforced explicitly in Linux. But, here we impose - * a limit of (STACK_SIZE - GUARD_SIZE) on the stack - * region, since we can do this with our mmap. - * - * Our mmap with MAP_STACK takes addr as the maximum - * downsize limit on BOS, and as len the max size of - * the region. It then maps the top SGROWSIZ bytes, - * and auto grows the region down, up to the limit - * in addr. - * - * If we don't use the MAP_STACK option, the effect - * of this code is to allocate a stack region of a - * fixed size of (STACK_SIZE - GUARD_SIZE). - */ - - if ((caddr_t)PTRIN(addr) + len > p->p_vmspace->vm_maxsaddr) { - /* - * Some Linux apps will attempt to mmap - * thread stacks near the top of their - * address space. If their TOS is greater - * than vm_maxsaddr, vm_map_growstack() - * will confuse the thread stack with the - * process stack and deliver a SEGV if they - * attempt to grow the thread stack past their - * current stacksize rlimit. To avoid this, - * adjust vm_maxsaddr upwards to reflect - * the current stacksize rlimit rather - * than the maximum possible stacksize. - * It would be better to adjust the - * mmap'ed region, but some apps do not check - * mmap's return value. - */ - PROC_LOCK(p); - p->p_vmspace->vm_maxsaddr = (char *)LINUX32_USRSTACK - - lim_cur_proc(p, RLIMIT_STACK); - PROC_UNLOCK(p); - } - - /* - * This gives us our maximum stack size and a new BOS. - * If we're using VM_STACK, then mmap will just map - * the top SGROWSIZ bytes, and let the stack grow down - * to the limit at BOS. If we're not using VM_STACK - * we map the full stack, since we don't have a way - * to autogrow it. - */ - if (len > STACK_SIZE - GUARD_SIZE) { - bsd_args.addr = (caddr_t)PTRIN(addr); - bsd_args.len = len; - } else { - bsd_args.addr = (caddr_t)PTRIN(addr) - - (STACK_SIZE - GUARD_SIZE - len); - bsd_args.len = STACK_SIZE - GUARD_SIZE; - } - } else { - bsd_args.addr = (caddr_t)PTRIN(addr); - bsd_args.len = len; - } - bsd_args.pos = pos; - -#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 = sys_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 - return (error); -} - int linux_mprotect(struct thread *td, struct linux_mprotect_args *uap) { - struct mprotect_args bsd_args; - - bsd_args.addr = uap->addr; - bsd_args.len = uap->len; - bsd_args.prot = uap->prot; - if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC)) - bsd_args.prot |= PROT_READ | PROT_EXEC; - return (sys_mprotect(td, &bsd_args)); + + return (linux_mprotect_common(td, PTROUT(uap->addr), uap->len, uap->prot)); } int diff --git a/sys/compat/linux/linux_mmap.c b/sys/compat/linux/linux_mmap.c new file mode 100644 index 000000000000..870f0be418f6 --- /dev/null +++ b/sys/compat/linux/linux_mmap.c @@ -0,0 +1,257 @@ +/*- + * Copyright (c) 2004 Tim J. Robbins + * Copyright (c) 2002 Doug Rabson + * Copyright (c) 2000 Marcel Moolenaar + * Copyright (c) 1994-1995 Søren Schmidt + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer + * in this position and unchanged. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. The name of the author may not be used to endorse or promote products + * derived from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES + * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. + * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, + * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT + * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF + * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + * + * $FreeBSD$ + */ + +#include <sys/cdefs.h> +__FBSDID("$FreeBSD$"); + +#include <sys/capsicum.h> +#include <sys/file.h> +#include <sys/imgact.h> +#include <sys/ktr.h> +#include <sys/mman.h> +#include <sys/proc.h> +#include <sys/resourcevar.h> +#include <sys/sysent.h> +#include <sys/sysproto.h> + +#include <vm/pmap.h> +#include <vm/vm_map.h> + +#include <compat/linux/linux_emul.h> +#include <compat/linux/linux_mmap.h> +#include <compat/linux/linux_persona.h> +#include <compat/linux/linux_util.h> + + +#define STACK_SIZE (2 * 1024 * 1024) +#define GUARD_SIZE (4 * PAGE_SIZE) + +#if defined(__amd64__) +static void linux_fixup_prot(struct thread *td, int *prot); +#endif + + +int +linux_mmap_common(struct thread *td, uintptr_t addr, size_t len, int prot, + int flags, int fd, off_t pos) +{ + struct proc *p = td->td_proc; + struct vmspace *vms = td->td_proc->p_vmspace; + struct mmap_args /* { + caddr_t addr; + size_t len; + int prot; + int flags; + int fd; + off_t pos; + } */ bsd_args; + int error; + struct file *fp; + + cap_rights_t rights; + LINUX_CTR6(mmap2, "0x%lx, %ld, %ld, 0x%08lx, %ld, 0x%lx", + addr, len, prot, flags, fd, pos); + + error = 0; + bsd_args.flags = 0; + fp = NULL; + + /* + * Linux mmap(2): + * You must specify exactly one of MAP_SHARED and MAP_PRIVATE + */ + if (!((flags & LINUX_MAP_SHARED) ^ (flags & LINUX_MAP_PRIVATE))) + return (EINVAL); + + if (flags & LINUX_MAP_SHARED) + bsd_args.flags |= MAP_SHARED; + if (flags & LINUX_MAP_PRIVATE) + bsd_args.flags |= MAP_PRIVATE; + if (flags & LINUX_MAP_FIXED) + bsd_args.flags |= MAP_FIXED; + if (flags & LINUX_MAP_ANON) { + /* Enforce pos to be on page boundary, then ignore. */ + if ((pos & PAGE_MASK) != 0) + return (EINVAL); + pos = 0; + bsd_args.flags |= MAP_ANON; + } else + bsd_args.flags |= MAP_NOSYNC; + if (flags & LINUX_MAP_GROWSDOWN) + bsd_args.flags |= MAP_STACK; + + /* + * PROT_READ, PROT_WRITE, or PROT_EXEC implies PROT_READ and PROT_EXEC + * on Linux/i386 if the binary requires executable stack. + * We do this only for IA32 emulation as on native i386 this is does not + * make sense without PAE. + * + * XXX. Linux checks that the file system is not mounted with noexec. + */ + bsd_args.prot = prot; +#if defined(__amd64__) + linux_fixup_prot(td, &bsd_args.prot); +#endif + + /* Linux does not check file descriptor when MAP_ANONYMOUS is set. */ + bsd_args.fd = (bsd_args.flags & MAP_ANON) ? -1 : fd; + if (bsd_args.fd != -1) { + /* + * Linux follows Solaris mmap(2) description: + * The file descriptor fildes is opened with + * read permission, regardless of the + * protection options specified. + */ + + error = fget(td, bsd_args.fd, + cap_rights_init(&rights, CAP_MMAP), &fp); + if (error != 0) + return (error); + if (fp->f_type != DTYPE_VNODE) { + fdrop(fp, td); + return (EINVAL); + } + + /* Linux mmap() just fails for O_WRONLY files */ + if (!(fp->f_flag & FREAD)) { + fdrop(fp, td); + return (EACCES); + } + + fdrop(fp, td); + } + + if (flags & LINUX_MAP_GROWSDOWN) { + /* + * The Linux MAP_GROWSDOWN option does not limit auto + * growth of the region. Linux mmap with this option + * takes as addr the initial BOS, and as len, the initial + * region size. It can then grow down from addr without + * limit. However, Linux threads has an implicit internal + * limit to stack size of STACK_SIZE. Its just not + * enforced explicitly in Linux. But, here we impose + * a limit of (STACK_SIZE - GUARD_SIZE) on the stack + * region, since we can do this with our mmap. + * + * Our mmap with MAP_STACK takes addr as the maximum + * downsize limit on BOS, and as len the max size of + * the region. It then maps the top SGROWSIZ bytes, + * and auto grows the region down, up to the limit + * in addr. + * + * If we don't use the MAP_STACK option, the effect + * of this code is to allocate a stack region of a + * fixed size of (STACK_SIZE - GUARD_SIZE). + */ + + if ((caddr_t)addr + len > vms->vm_maxsaddr) { + /* + * Some Linux apps will attempt to mmap + * thread stacks near the top of their + * address space. If their TOS is greater + * than vm_maxsaddr, vm_map_growstack() + * will confuse the thread stack with the + * process stack and deliver a SEGV if they + * attempt to grow the thread stack past their + * current stacksize rlimit. To avoid this, + * adjust vm_maxsaddr upwards to reflect + * the current stacksize rlimit rather + * than the maximum possible stacksize. + * It would be better to adjust the + * mmap'ed region, but some apps do not check + * mmap's return value. + */ + PROC_LOCK(p); + vms->vm_maxsaddr = (char *)p->p_sysent->sv_usrstack - + lim_cur_proc(p, RLIMIT_STACK); + PROC_UNLOCK(p); + } + + /* + * This gives us our maximum stack size and a new BOS. + * If we're using VM_STACK, then mmap will just map + * the top SGROWSIZ bytes, and let the stack grow down + * to the limit at BOS. If we're not using VM_STACK + * we map the full stack, since we don't have a way + * to autogrow it. + */ + if (len > STACK_SIZE - GUARD_SIZE) { + bsd_args.addr = (caddr_t)addr; + bsd_args.len = len; + } else { + bsd_args.addr = (caddr_t)addr - + (STACK_SIZE - GUARD_SIZE - len); + bsd_args.len = STACK_SIZE - GUARD_SIZE; + } + } else { + bsd_args.addr = (caddr_t)addr; + bsd_args.len = len; + } + bsd_args.pos = pos; + + error = sys_mmap(td, &bsd_args); + + LINUX_CTR2(mmap2, "return: %d (%p)", error, td->td_retval[0]); + + return (error); +} + +int +linux_mprotect_common(struct thread *td, uintptr_t addr, size_t len, int prot) +{ + struct mprotect_args bsd_args; + + bsd_args.addr = (void *)addr; + bsd_args.len = len; + bsd_args.prot = prot; + +#if defined(__amd64__) + linux_fixup_prot(td, &bsd_args.prot); +#endif + return (sys_mprotect(td, &bsd_args)); +} + +#if defined(__amd64__) +static void +linux_fixup_prot(struct thread *td, int *prot) +{ + struct linux_pemuldata *pem; + + if (SV_PROC_FLAG(td->td_proc, SV_ILP32) && *prot & PROT_READ) { + pem = pem_find(td->td_proc); + if (pem->persona & LINUX_READ_IMPLIES_EXEC) + *prot |= PROT_EXEC; + } + +} +#endif diff --git a/sys/compat/linux/linux_mmap.h b/sys/compat/linux/linux_mmap.h new file mode 100644 index 000000000000..a27d99d989d5 --- /dev/null +++ b/sys/compat/linux/linux_mmap.h @@ -0,0 +1,49 @@ +/*- + * Copyright (c) 2004 Tim J. Robbins + * Copyright (c) 2002 Doug Rabson + * Copyright (c) 2000 Marcel Moolenaar + * Copyright (c) 1994-1995 Søren Schmidt + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer + * in this position and unchanged. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. The name of the author may not be used to endorse or promote products + * derived from this software without specific prior written permission. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES + * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. + * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, + * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT + * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF + * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + * + * $FreeBSD$ + */ + +#ifndef _LINUX_MMAP_H_ +#define _LINUX_MMAP_H_ + +/* mmap options */ +#define LINUX_MAP_SHARED 0x0001 +#define LINUX_MAP_PRIVATE 0x0002 +#define LINUX_MAP_FIXED 0x0010 +#define LINUX_MAP_ANON 0x0020 +#define LINUX_MAP_GROWSDOWN 0x0100 + + +int linux_mmap_common(struct thread *, uintptr_t, size_t, int, int, + int, off_t); +int linux_mprotect_common(struct thread *, uintptr_t, size_t, int); + +#endif /* _LINUX_MMAP_H_ */ diff --git a/sys/i386/linux/linux.h b/sys/i386/linux/linux.h index 36b2084cd9f9..42e836ca182e 100644 --- a/sys/i386/linux/linux.h +++ b/sys/i386/linux/linux.h @@ -140,13 +140,6 @@ struct l_rlimit { l_ulong rlim_max; }; -/* mmap options */ -#define LINUX_MAP_SHARED 0x0001 -#define LINUX_MAP_PRIVATE 0x0002 -#define LINUX_MAP_FIXED 0x0010 -#define LINUX_MAP_ANON 0x0020 -#define LINUX_MAP_GROWSDOWN 0x0100 - struct l_mmap_argv { l_uintptr_t addr; l_size_t len; diff --git a/sys/i386/linux/linux_machdep.c b/sys/i386/linux/linux_machdep.c index 4fb5785da4db..4b4b88609428 100644 --- a/sys/i386/linux/linux_machdep.c +++ b/sys/i386/linux/linux_machdep.c @@ -65,6 +65,7 @@ __FBSDID("$FreeBSD$"); #include <i386/linux/linux_proto.h> #include <compat/linux/linux_ipc.h> #include <compat/linux/linux_misc.h> +#include <compat/linux/linux_mmap.h> #include <compat/linux/linux_signal.h> #include <compat/linux/linux_util.h> #include <compat/linux/linux_emul.h> @@ -95,10 +96,6 @@ struct l_old_select_argv { struct l_timeval *timeout; }; -static int linux_mmap_common(struct thread *td, l_uintptr_t addr, - l_size_t len, l_int prot, l_int flags, l_int fd, - l_loff_t pos); - int linux_execve(struct thread *td, struct linux_execve_args *args) @@ -340,9 +337,6 @@ linux_set_upcall_kse(struct thread *td, register_t stack) return (0); } -#define STACK_SIZE (2 * 1024 * 1024) -#define GUARD_SIZE (4 * PAGE_SIZE) - int linux_mmap2(struct thread *td, struct linux_mmap2_args *args) { @@ -381,187 +375,11 @@ linux_mmap(struct thread *td, struct linux_mmap_args *args) (uint32_t)linux_args.pgoff)); } -static int -linux_mmap_common(struct thread *td, l_uintptr_t addr, l_size_t len, l_int prot, - l_int flags, l_int fd, l_loff_t pos) -{ - struct proc *p = td->td_proc; - struct mmap_args /* { - caddr_t addr; - size_t len; - int prot; - int flags; - int fd; - long pad; - off_t pos; - } */ bsd_args; - int error; - struct file *fp; - cap_rights_t rights; - - error = 0; - bsd_args.flags = 0; - fp = NULL; - - /* - * Linux mmap(2): - * You must specify exactly one of MAP_SHARED and MAP_PRIVATE - */ - if (!((flags & LINUX_MAP_SHARED) ^ (flags & LINUX_MAP_PRIVATE))) - return (EINVAL); - - if (flags & LINUX_MAP_SHARED) - bsd_args.flags |= MAP_SHARED; - if (flags & LINUX_MAP_PRIVATE) - bsd_args.flags |= MAP_PRIVATE; - if (flags & LINUX_MAP_FIXED) - bsd_args.flags |= MAP_FIXED; - if (flags & LINUX_MAP_ANON) { - /* Enforce pos to be on page boundary, then ignore. */ - if ((pos & PAGE_MASK) != 0) - return (EINVAL); - pos = 0; - bsd_args.flags |= MAP_ANON; - } else - bsd_args.flags |= MAP_NOSYNC; - if (flags & LINUX_MAP_GROWSDOWN) - bsd_args.flags |= MAP_STACK; - - /* - * PROT_READ, PROT_WRITE, or PROT_EXEC implies PROT_READ and PROT_EXEC - * on Linux/i386. We do this to ensure maximum compatibility. - * Linux/ia64 does the same in i386 emulation mode. - */ - bsd_args.prot = prot; - if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC)) - bsd_args.prot |= PROT_READ | PROT_EXEC; - - /* Linux does not check file descriptor when MAP_ANONYMOUS is set. */ - bsd_args.fd = (bsd_args.flags & MAP_ANON) ? -1 : fd; - if (bsd_args.fd != -1) { - /* - * Linux follows Solaris mmap(2) description: - * The file descriptor fildes is opened with - * read permission, regardless of the - * protection options specified. - * - * Checking just CAP_MMAP is fine here, since the real work - * is done in the FreeBSD mmap(). - */ - - error = fget(td, bsd_args.fd, - cap_rights_init(&rights, CAP_MMAP), &fp); - if (error != 0) - return (error); - if (fp->f_type != DTYPE_VNODE) { - fdrop(fp, td); - return (EINVAL); - } - - /* Linux mmap() just fails for O_WRONLY files */ - if (!(fp->f_flag & FREAD)) { - fdrop(fp, td); - return (EACCES); - } - - fdrop(fp, td); - } - - if (flags & LINUX_MAP_GROWSDOWN) { - /* - * The Linux MAP_GROWSDOWN option does not limit auto - * growth of the region. Linux mmap with this option - * takes as addr the inital BOS, and as len, the initial - * region size. It can then grow down from addr without - * limit. However, linux threads has an implicit internal - * limit to stack size of STACK_SIZE. Its just not - * enforced explicitly in linux. But, here we impose - * a limit of (STACK_SIZE - GUARD_SIZE) on the stack - * region, since we can do this with our mmap. - * - * Our mmap with MAP_STACK takes addr as the maximum - * downsize limit on BOS, and as len the max size of - * the region. It them maps the top SGROWSIZ bytes, - * and auto grows the region down, up to the limit - * in addr. - * - * If we don't use the MAP_STACK option, the effect - * of this code is to allocate a stack region of a - * fixed size of (STACK_SIZE - GUARD_SIZE). - */ - - if ((caddr_t)PTRIN(addr) + len > p->p_vmspace->vm_maxsaddr) { - /* - * Some linux apps will attempt to mmap - * thread stacks near the top of their - * address space. If their TOS is greater - * than vm_maxsaddr, vm_map_growstack() - * will confuse the thread stack with the - * process stack and deliver a SEGV if they - * attempt to grow the thread stack past their - * current stacksize rlimit. To avoid this, - * adjust vm_maxsaddr upwards to reflect - * the current stacksize rlimit rather - * than the maximum possible stacksize. - * It would be better to adjust the - * mmap'ed region, but some apps do not check - * mmap's return value. - */ - PROC_LOCK(p); - p->p_vmspace->vm_maxsaddr = (char *)USRSTACK - - lim_cur_proc(p, RLIMIT_STACK); - PROC_UNLOCK(p); - } - - /* - * This gives us our maximum stack size and a new BOS. - * If we're using VM_STACK, then mmap will just map - * the top SGROWSIZ bytes, and let the stack grow down - * to the limit at BOS. If we're not using VM_STACK - * we map the full stack, since we don't have a way - * to autogrow it. - */ - if (len > STACK_SIZE - GUARD_SIZE) { - bsd_args.addr = (caddr_t)PTRIN(addr); - bsd_args.len = len; - } else { - bsd_args.addr = (caddr_t)PTRIN(addr) - - (STACK_SIZE - GUARD_SIZE - len); - bsd_args.len = STACK_SIZE - GUARD_SIZE; - } - } else { - bsd_args.addr = (caddr_t)PTRIN(addr); - bsd_args.len = len; - } - bsd_args.pos = pos; - -#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 = sys_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 - return (error); -} - int linux_mprotect(struct thread *td, struct linux_mprotect_args *uap) { - struct mprotect_args bsd_args; - - bsd_args.addr = uap->addr; - bsd_args.len = uap->len; - bsd_args.prot = uap->prot; - if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC)) - bsd_args.prot |= PROT_READ | PROT_EXEC; - return (sys_mprotect(td, &bsd_args)); + + return (linux_mprotect_common(td, PTROUT(uap->addr), uap->len, uap->prot)); } int diff --git a/sys/modules/linux/Makefile b/sys/modules/linux/Makefile index f2629723d404..d1707c92e91e 100644 --- a/sys/modules/linux/Makefile +++ b/sys/modules/linux/Makefile @@ -30,7 +30,7 @@ SRCS+= opt_apic.h OBJS= ${VDSO}.so .if ${MACHINE_CPUARCH} == "i386" -SRCS+= linux_ptrace.c imgact_linux.c linux_util.c linux_mib.c \ +SRCS+= linux_ptrace.c imgact_linux.c linux_util.c linux_mib.c linux_mmap.c \ linux_emul.c opt_cpu.h linux.c .endif diff --git a/sys/modules/linux_common/Makefile b/sys/modules/linux_common/Makefile index 91449f774ef5..2301796230b0 100644 --- a/sys/modules/linux_common/Makefile +++ b/sys/modules/linux_common/Makefile @@ -3,7 +3,7 @@ .PATH: ${.CURDIR}/../../compat/linux KMOD= linux_common -SRCS= linux_common.c linux_mib.c linux_util.c linux_emul.c \ +SRCS= linux_common.c linux_mib.c linux_mmap.c linux_util.c linux_emul.c \ linux.c opt_compat.h device_if.h vnode_if.h bus_if.h EXPORT_SYMS= |