freebsd-dev/sys/kern/uipc_syscalls.c

1758 lines
36 KiB
C
Raw Normal View History

/*-
1994-05-24 10:09:53 +00:00
* Copyright (c) 1982, 1986, 1989, 1990, 1993
* The Regents of the University of California. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* @(#)uipc_syscalls.c 8.4 (Berkeley) 2/21/94
*/
2003-06-11 00:56:59 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_capsicum.h"
#include "opt_inet.h"
#include "opt_inet6.h"
#include "opt_compat.h"
#include "opt_ktrace.h"
1994-05-24 10:09:53 +00:00
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/capsicum.h>
#include <sys/kernel.h>
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/sysproto.h>
#include <sys/malloc.h>
1994-05-24 10:09:53 +00:00
#include <sys/filedesc.h>
#include <sys/proc.h>
#include <sys/filio.h>
#include <sys/jail.h>
1994-05-24 10:09:53 +00:00
#include <sys/mbuf.h>
#include <sys/protosw.h>
Switch the vm_object mutex to be a rwlock. This will enable in the future further optimizations where the vm_object lock will be held in read mode most of the time the page cache resident pool of pages are accessed for reading purposes. The change is mostly mechanical but few notes are reported: * The KPI changes as follow: - VM_OBJECT_LOCK() -> VM_OBJECT_WLOCK() - VM_OBJECT_TRYLOCK() -> VM_OBJECT_TRYWLOCK() - VM_OBJECT_UNLOCK() -> VM_OBJECT_WUNLOCK() - VM_OBJECT_LOCK_ASSERT(MA_OWNED) -> VM_OBJECT_ASSERT_WLOCKED() (in order to avoid visibility of implementation details) - The read-mode operations are added: VM_OBJECT_RLOCK(), VM_OBJECT_TRYRLOCK(), VM_OBJECT_RUNLOCK(), VM_OBJECT_ASSERT_RLOCKED(), VM_OBJECT_ASSERT_LOCKED() * The vm/vm_pager.h namespace pollution avoidance (forcing requiring sys/mutex.h in consumers directly to cater its inlining functions using VM_OBJECT_LOCK()) imposes that all the vm/vm_pager.h consumers now must include also sys/rwlock.h. * zfs requires a quite convoluted fix to include FreeBSD rwlocks into the compat layer because the name clash between FreeBSD and solaris versions must be avoided. At this purpose zfs redefines the vm_object locking functions directly, isolating the FreeBSD components in specific compat stubs. The KPI results heavilly broken by this commit. Thirdy part ports must be updated accordingly (I can think off-hand of VirtualBox, for example). Sponsored by: EMC / Isilon storage division Reviewed by: jeff Reviewed by: pjd (ZFS specific review) Discussed with: alc Tested by: pho
2013-03-09 02:32:23 +00:00
#include <sys/rwlock.h>
1994-05-24 10:09:53 +00:00
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/syscallsubr.h>
1994-05-24 10:09:53 +00:00
#ifdef KTRACE
#include <sys/ktrace.h>
#endif
#ifdef COMPAT_FREEBSD32
#include <compat/freebsd32/freebsd32_util.h>
#endif
#include <net/vnet.h>
#include <security/audit/audit.h>
#include <security/mac/mac_framework.h>
/*
* Flags for accept1() and kern_accept4(), in addition to SOCK_CLOEXEC
* and SOCK_NONBLOCK.
*/
#define ACCEPT4_INHERIT 0x1
#define ACCEPT4_COMPAT 0x2
2002-03-19 21:25:46 +00:00
static int sendit(struct thread *td, int s, struct msghdr *mp, int flags);
static int recvit(struct thread *td, int s, struct msghdr *mp, void *namelenp);
static int accept1(struct thread *td, int s, struct sockaddr *uname,
socklen_t *anamelen, int flags);
2002-03-19 21:25:46 +00:00
static int getsockname1(struct thread *td, struct getsockname_args *uap,
int compat);
2002-03-19 21:25:46 +00:00
static int getpeername1(struct thread *td, struct getpeername_args *uap,
int compat);
/*
Merge Capsicum overhaul: - Capability is no longer separate descriptor type. Now every descriptor has set of its own capability rights. - The cap_new(2) system call is left, but it is no longer documented and should not be used in new code. - The new syscall cap_rights_limit(2) should be used instead of cap_new(2), which limits capability rights of the given descriptor without creating a new one. - The cap_getrights(2) syscall is renamed to cap_rights_get(2). - If CAP_IOCTL capability right is present we can further reduce allowed ioctls list with the new cap_ioctls_limit(2) syscall. List of allowed ioctls can be retrived with cap_ioctls_get(2) syscall. - If CAP_FCNTL capability right is present we can further reduce fcntls that can be used with the new cap_fcntls_limit(2) syscall and retrive them with cap_fcntls_get(2). - To support ioctl and fcntl white-listing the filedesc structure was heavly modified. - The audit subsystem, kdump and procstat tools were updated to recognize new syscalls. - Capability rights were revised and eventhough I tried hard to provide backward API and ABI compatibility there are some incompatible changes that are described in detail below: CAP_CREATE old behaviour: - Allow for openat(2)+O_CREAT. - Allow for linkat(2). - Allow for symlinkat(2). CAP_CREATE new behaviour: - Allow for openat(2)+O_CREAT. Added CAP_LINKAT: - Allow for linkat(2). ABI: Reuses CAP_RMDIR bit. - Allow to be target for renameat(2). Added CAP_SYMLINKAT: - Allow for symlinkat(2). Removed CAP_DELETE. Old behaviour: - Allow for unlinkat(2) when removing non-directory object. - Allow to be source for renameat(2). Removed CAP_RMDIR. Old behaviour: - Allow for unlinkat(2) when removing directory. Added CAP_RENAMEAT: - Required for source directory for the renameat(2) syscall. Added CAP_UNLINKAT (effectively it replaces CAP_DELETE and CAP_RMDIR): - Allow for unlinkat(2) on any object. - Required if target of renameat(2) exists and will be removed by this call. Removed CAP_MAPEXEC. CAP_MMAP old behaviour: - Allow for mmap(2) with any combination of PROT_NONE, PROT_READ and PROT_WRITE. CAP_MMAP new behaviour: - Allow for mmap(2)+PROT_NONE. Added CAP_MMAP_R: - Allow for mmap(PROT_READ). Added CAP_MMAP_W: - Allow for mmap(PROT_WRITE). Added CAP_MMAP_X: - Allow for mmap(PROT_EXEC). Added CAP_MMAP_RW: - Allow for mmap(PROT_READ | PROT_WRITE). Added CAP_MMAP_RX: - Allow for mmap(PROT_READ | PROT_EXEC). Added CAP_MMAP_WX: - Allow for mmap(PROT_WRITE | PROT_EXEC). Added CAP_MMAP_RWX: - Allow for mmap(PROT_READ | PROT_WRITE | PROT_EXEC). Renamed CAP_MKDIR to CAP_MKDIRAT. Renamed CAP_MKFIFO to CAP_MKFIFOAT. Renamed CAP_MKNODE to CAP_MKNODEAT. CAP_READ old behaviour: - Allow pread(2). - Disallow read(2), readv(2) (if there is no CAP_SEEK). CAP_READ new behaviour: - Allow read(2), readv(2). - Disallow pread(2) (CAP_SEEK was also required). CAP_WRITE old behaviour: - Allow pwrite(2). - Disallow write(2), writev(2) (if there is no CAP_SEEK). CAP_WRITE new behaviour: - Allow write(2), writev(2). - Disallow pwrite(2) (CAP_SEEK was also required). Added convinient defines: #define CAP_PREAD (CAP_SEEK | CAP_READ) #define CAP_PWRITE (CAP_SEEK | CAP_WRITE) #define CAP_MMAP_R (CAP_MMAP | CAP_SEEK | CAP_READ) #define CAP_MMAP_W (CAP_MMAP | CAP_SEEK | CAP_WRITE) #define CAP_MMAP_X (CAP_MMAP | CAP_SEEK | 0x0000000000000008ULL) #define CAP_MMAP_RW (CAP_MMAP_R | CAP_MMAP_W) #define CAP_MMAP_RX (CAP_MMAP_R | CAP_MMAP_X) #define CAP_MMAP_WX (CAP_MMAP_W | CAP_MMAP_X) #define CAP_MMAP_RWX (CAP_MMAP_R | CAP_MMAP_W | CAP_MMAP_X) #define CAP_RECV CAP_READ #define CAP_SEND CAP_WRITE #define CAP_SOCK_CLIENT \ (CAP_CONNECT | CAP_GETPEERNAME | CAP_GETSOCKNAME | CAP_GETSOCKOPT | \ CAP_PEELOFF | CAP_RECV | CAP_SEND | CAP_SETSOCKOPT | CAP_SHUTDOWN) #define CAP_SOCK_SERVER \ (CAP_ACCEPT | CAP_BIND | CAP_GETPEERNAME | CAP_GETSOCKNAME | \ CAP_GETSOCKOPT | CAP_LISTEN | CAP_PEELOFF | CAP_RECV | CAP_SEND | \ CAP_SETSOCKOPT | CAP_SHUTDOWN) Added defines for backward API compatibility: #define CAP_MAPEXEC CAP_MMAP_X #define CAP_DELETE CAP_UNLINKAT #define CAP_MKDIR CAP_MKDIRAT #define CAP_RMDIR CAP_UNLINKAT #define CAP_MKFIFO CAP_MKFIFOAT #define CAP_MKNOD CAP_MKNODAT #define CAP_SOCK_ALL (CAP_SOCK_CLIENT | CAP_SOCK_SERVER) Sponsored by: The FreeBSD Foundation Reviewed by: Christoph Mallon <christoph.mallon@gmx.de> Many aspects discussed with: rwatson, benl, jonathan ABI compatibility discussed with: kib
2013-03-02 00:53:12 +00:00
* Convert a user file descriptor to a kernel file entry and check if required
* capability rights are present.
* A reference on the file entry is held upon returning.
*/
int
getsock_cap(struct thread *td, int fd, cap_rights_t *rightsp,
struct file **fpp, u_int *fflagp)
{
struct file *fp;
int error;
error = fget_unlocked(td->td_proc->p_fd, fd, rightsp, &fp, NULL);
Merge Capsicum overhaul: - Capability is no longer separate descriptor type. Now every descriptor has set of its own capability rights. - The cap_new(2) system call is left, but it is no longer documented and should not be used in new code. - The new syscall cap_rights_limit(2) should be used instead of cap_new(2), which limits capability rights of the given descriptor without creating a new one. - The cap_getrights(2) syscall is renamed to cap_rights_get(2). - If CAP_IOCTL capability right is present we can further reduce allowed ioctls list with the new cap_ioctls_limit(2) syscall. List of allowed ioctls can be retrived with cap_ioctls_get(2) syscall. - If CAP_FCNTL capability right is present we can further reduce fcntls that can be used with the new cap_fcntls_limit(2) syscall and retrive them with cap_fcntls_get(2). - To support ioctl and fcntl white-listing the filedesc structure was heavly modified. - The audit subsystem, kdump and procstat tools were updated to recognize new syscalls. - Capability rights were revised and eventhough I tried hard to provide backward API and ABI compatibility there are some incompatible changes that are described in detail below: CAP_CREATE old behaviour: - Allow for openat(2)+O_CREAT. - Allow for linkat(2). - Allow for symlinkat(2). CAP_CREATE new behaviour: - Allow for openat(2)+O_CREAT. Added CAP_LINKAT: - Allow for linkat(2). ABI: Reuses CAP_RMDIR bit. - Allow to be target for renameat(2). Added CAP_SYMLINKAT: - Allow for symlinkat(2). Removed CAP_DELETE. Old behaviour: - Allow for unlinkat(2) when removing non-directory object. - Allow to be source for renameat(2). Removed CAP_RMDIR. Old behaviour: - Allow for unlinkat(2) when removing directory. Added CAP_RENAMEAT: - Required for source directory for the renameat(2) syscall. Added CAP_UNLINKAT (effectively it replaces CAP_DELETE and CAP_RMDIR): - Allow for unlinkat(2) on any object. - Required if target of renameat(2) exists and will be removed by this call. Removed CAP_MAPEXEC. CAP_MMAP old behaviour: - Allow for mmap(2) with any combination of PROT_NONE, PROT_READ and PROT_WRITE. CAP_MMAP new behaviour: - Allow for mmap(2)+PROT_NONE. Added CAP_MMAP_R: - Allow for mmap(PROT_READ). Added CAP_MMAP_W: - Allow for mmap(PROT_WRITE). Added CAP_MMAP_X: - Allow for mmap(PROT_EXEC). Added CAP_MMAP_RW: - Allow for mmap(PROT_READ | PROT_WRITE). Added CAP_MMAP_RX: - Allow for mmap(PROT_READ | PROT_EXEC). Added CAP_MMAP_WX: - Allow for mmap(PROT_WRITE | PROT_EXEC). Added CAP_MMAP_RWX: - Allow for mmap(PROT_READ | PROT_WRITE | PROT_EXEC). Renamed CAP_MKDIR to CAP_MKDIRAT. Renamed CAP_MKFIFO to CAP_MKFIFOAT. Renamed CAP_MKNODE to CAP_MKNODEAT. CAP_READ old behaviour: - Allow pread(2). - Disallow read(2), readv(2) (if there is no CAP_SEEK). CAP_READ new behaviour: - Allow read(2), readv(2). - Disallow pread(2) (CAP_SEEK was also required). CAP_WRITE old behaviour: - Allow pwrite(2). - Disallow write(2), writev(2) (if there is no CAP_SEEK). CAP_WRITE new behaviour: - Allow write(2), writev(2). - Disallow pwrite(2) (CAP_SEEK was also required). Added convinient defines: #define CAP_PREAD (CAP_SEEK | CAP_READ) #define CAP_PWRITE (CAP_SEEK | CAP_WRITE) #define CAP_MMAP_R (CAP_MMAP | CAP_SEEK | CAP_READ) #define CAP_MMAP_W (CAP_MMAP | CAP_SEEK | CAP_WRITE) #define CAP_MMAP_X (CAP_MMAP | CAP_SEEK | 0x0000000000000008ULL) #define CAP_MMAP_RW (CAP_MMAP_R | CAP_MMAP_W) #define CAP_MMAP_RX (CAP_MMAP_R | CAP_MMAP_X) #define CAP_MMAP_WX (CAP_MMAP_W | CAP_MMAP_X) #define CAP_MMAP_RWX (CAP_MMAP_R | CAP_MMAP_W | CAP_MMAP_X) #define CAP_RECV CAP_READ #define CAP_SEND CAP_WRITE #define CAP_SOCK_CLIENT \ (CAP_CONNECT | CAP_GETPEERNAME | CAP_GETSOCKNAME | CAP_GETSOCKOPT | \ CAP_PEELOFF | CAP_RECV | CAP_SEND | CAP_SETSOCKOPT | CAP_SHUTDOWN) #define CAP_SOCK_SERVER \ (CAP_ACCEPT | CAP_BIND | CAP_GETPEERNAME | CAP_GETSOCKNAME | \ CAP_GETSOCKOPT | CAP_LISTEN | CAP_PEELOFF | CAP_RECV | CAP_SEND | \ CAP_SETSOCKOPT | CAP_SHUTDOWN) Added defines for backward API compatibility: #define CAP_MAPEXEC CAP_MMAP_X #define CAP_DELETE CAP_UNLINKAT #define CAP_MKDIR CAP_MKDIRAT #define CAP_RMDIR CAP_UNLINKAT #define CAP_MKFIFO CAP_MKFIFOAT #define CAP_MKNOD CAP_MKNODAT #define CAP_SOCK_ALL (CAP_SOCK_CLIENT | CAP_SOCK_SERVER) Sponsored by: The FreeBSD Foundation Reviewed by: Christoph Mallon <christoph.mallon@gmx.de> Many aspects discussed with: rwatson, benl, jonathan ABI compatibility discussed with: kib
2013-03-02 00:53:12 +00:00
if (error != 0)
return (error);
if (fp->f_type != DTYPE_SOCKET) {
fdrop(fp, td);
return (ENOTSOCK);
}
if (fflagp != NULL)
*fflagp = fp->f_flag;
*fpp = fp;
return (0);
}
1994-05-24 10:09:53 +00:00
/*
* System call interface to the socket abstraction.
*/
#if defined(COMPAT_43)
1994-05-24 10:09:53 +00:00
#define COMPAT_OLDSOCK
#endif
int
sys_socket(td, uap)
struct thread *td;
struct socket_args /* {
int domain;
int type;
int protocol;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct socket *so;
struct file *fp;
int fd, error, type, oflag, fflag;
1994-05-24 10:09:53 +00:00
AUDIT_ARG_SOCKET(uap->domain, uap->type, uap->protocol);
type = uap->type;
oflag = 0;
fflag = 0;
if ((type & SOCK_CLOEXEC) != 0) {
type &= ~SOCK_CLOEXEC;
oflag |= O_CLOEXEC;
}
if ((type & SOCK_NONBLOCK) != 0) {
type &= ~SOCK_NONBLOCK;
fflag |= FNONBLOCK;
}
#ifdef MAC
error = mac_socket_check_create(td->td_ucred, uap->domain, type,
uap->protocol);
if (error != 0)
return (error);
#endif
error = falloc(td, &fp, &fd, oflag);
if (error != 0)
return (error);
/* An extra reference on `fp' has been held for us by falloc(). */
error = socreate(uap->domain, &so, type, uap->protocol,
td->td_ucred, td);
if (error != 0) {
fdclose(td, fp, fd);
1994-05-24 10:09:53 +00:00
} else {
finit(fp, FREAD | FWRITE | fflag, DTYPE_SOCKET, so, &socketops);
if ((fflag & FNONBLOCK) != 0)
(void) fo_ioctl(fp, FIONBIO, &fflag, td->td_ucred, td);
td->td_retval[0] = fd;
1994-05-24 10:09:53 +00:00
}
fdrop(fp, td);
1994-05-24 10:09:53 +00:00
return (error);
}
/* ARGSUSED */
int
sys_bind(td, uap)
struct thread *td;
struct bind_args /* {
int s;
caddr_t name;
int namelen;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct sockaddr *sa;
1994-05-24 10:09:53 +00:00
int error;
error = getsockaddr(&sa, uap->name, uap->namelen);
if (error == 0) {
error = kern_bindat(td, AT_FDCWD, uap->s, sa);
free(sa, M_SONAME);
}
return (error);
}
int
kern_bindat(struct thread *td, int dirfd, int fd, struct sockaddr *sa)
{
struct socket *so;
struct file *fp;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
int error;
AUDIT_ARG_FD(fd);
AUDIT_ARG_SOCKADDR(td, dirfd, sa);
error = getsock_cap(td, fd, cap_rights_init(&rights, CAP_BIND),
&fp, NULL);
if (error != 0)
return (error);
so = fp->f_data;
#ifdef KTRACE
if (KTRPOINT(td, KTR_STRUCT))
ktrsockaddr(sa);
#endif
#ifdef MAC
error = mac_socket_check_bind(td->td_ucred, so, sa);
if (error == 0) {
#endif
if (dirfd == AT_FDCWD)
error = sobind(so, sa, td);
else
error = sobindat(dirfd, so, sa, td);
#ifdef MAC
}
#endif
fdrop(fp, td);
1994-05-24 10:09:53 +00:00
return (error);
}
/* ARGSUSED */
int
sys_bindat(td, uap)
struct thread *td;
struct bindat_args /* {
int fd;
int s;
caddr_t name;
int namelen;
} */ *uap;
{
struct sockaddr *sa;
int error;
error = getsockaddr(&sa, uap->name, uap->namelen);
if (error == 0) {
error = kern_bindat(td, uap->fd, uap->s, sa);
free(sa, M_SONAME);
}
return (error);
}
1994-05-24 10:09:53 +00:00
/* ARGSUSED */
int
sys_listen(td, uap)
struct thread *td;
struct listen_args /* {
int s;
int backlog;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
2002-01-09 02:47:00 +00:00
struct socket *so;
struct file *fp;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
1994-05-24 10:09:53 +00:00
int error;
AUDIT_ARG_FD(uap->s);
error = getsock_cap(td, uap->s, cap_rights_init(&rights, CAP_LISTEN),
&fp, NULL);
if (error == 0) {
so = fp->f_data;
#ifdef MAC
error = mac_socket_check_listen(td->td_ucred, so);
if (error == 0)
#endif
error = solisten(so, uap->backlog, td);
fdrop(fp, td);
}
return(error);
1994-05-24 10:09:53 +00:00
}
/*
* accept1()
*/
static int
accept1(td, s, uname, anamelen, flags)
struct thread *td;
int s;
struct sockaddr *uname;
socklen_t *anamelen;
int flags;
{
struct sockaddr *name;
socklen_t namelen;
struct file *fp;
int error;
if (uname == NULL)
return (kern_accept4(td, s, NULL, NULL, flags, NULL));
error = copyin(anamelen, &namelen, sizeof (namelen));
if (error != 0)
return (error);
error = kern_accept4(td, s, &name, &namelen, flags, &fp);
if (error != 0)
return (error);
if (error == 0 && uname != NULL) {
#ifdef COMPAT_OLDSOCK
if (flags & ACCEPT4_COMPAT)
((struct osockaddr *)name)->sa_family =
name->sa_family;
#endif
error = copyout(name, uname, namelen);
}
if (error == 0)
error = copyout(&namelen, anamelen,
sizeof(namelen));
if (error != 0)
fdclose(td, fp, td->td_retval[0]);
fdrop(fp, td);
free(name, M_SONAME);
return (error);
}
int
kern_accept(struct thread *td, int s, struct sockaddr **name,
socklen_t *namelen, struct file **fp)
{
return (kern_accept4(td, s, name, namelen, ACCEPT4_INHERIT, fp));
}
int
kern_accept4(struct thread *td, int s, struct sockaddr **name,
socklen_t *namelen, int flags, struct file **fp)
1994-05-24 10:09:53 +00:00
{
struct file *headfp, *nfp = NULL;
struct sockaddr *sa = NULL;
struct socket *head, *so;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
u_int fflag;
pid_t pgid;
int error, fd, tmp;
1994-05-24 10:09:53 +00:00
if (name != NULL)
*name = NULL;
AUDIT_ARG_FD(s);
error = getsock_cap(td, s, cap_rights_init(&rights, CAP_ACCEPT),
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
&headfp, &fflag);
if (error != 0)
return (error);
head = headfp->f_data;
if ((head->so_options & SO_ACCEPTCONN) == 0) {
error = EINVAL;
goto done;
1994-05-24 10:09:53 +00:00
}
#ifdef MAC
error = mac_socket_check_accept(td->td_ucred, head);
if (error != 0)
goto done;
#endif
error = falloc(td, &nfp, &fd, (flags & SOCK_CLOEXEC) ? O_CLOEXEC : 0);
if (error != 0)
Integrate accept locking from rwatson_netperf, introducing a new global mutex, accept_mtx, which serializes access to the following fields across all sockets: so_qlen so_incqlen so_qstate so_comp so_incomp so_list so_head While providing only coarse granularity, this approach avoids lock order issues between sockets by avoiding ownership of the fields by a specific socket and its per-socket mutexes. While here, rewrite soclose(), sofree(), soaccept(), and sonewconn() to add assertions, close additional races and address lock order concerns. In particular: - Reorganize the optimistic concurrency behavior in accept1() to always allocate a file descriptor with falloc() so that if we do find a socket, we don't have to encounter the "Oh, there wasn't a socket" race that can occur if falloc() sleeps in the current code, which broke inbound accept() ordering, not to mention requiring backing out socket state changes in a way that raced with the protocol level. We may want to add a lockless read of the queue state if polling of empty queues proves to be important to optimize. - In accept1(), soref() the socket while holding the accept lock so that the socket cannot be free'd in a race with the protocol layer. Likewise in netgraph equivilents of the accept1() code. - In sonewconn(), loop waiting for the queue to be small enough to insert our new socket once we've committed to inserting it, or races can occur that cause the incomplete socket queue to overfill. In the previously implementation, it was sufficient to simply tested once since calling soabort() didn't release synchronization permitting another thread to insert a socket as we discard a previous one. - In soclose()/sofree()/et al, it is the responsibility of the caller to remove a socket from the incomplete connection queue before calling soabort(), which prevents soabort() from having to walk into the accept socket to release the socket from its queue, and avoids races when releasing the accept mutex to enter soabort(), permitting soabort() to avoid lock ordering issues with the caller. - Generally cluster accept queue related operations together throughout these functions in order to facilitate locking. Annotate new locking in socketvar.h.
2004-06-02 04:15:39 +00:00
goto done;
ACCEPT_LOCK();
if ((head->so_state & SS_NBIO) && TAILQ_EMPTY(&head->so_comp)) {
ACCEPT_UNLOCK();
error = EWOULDBLOCK;
goto noconnection;
Integrate accept locking from rwatson_netperf, introducing a new global mutex, accept_mtx, which serializes access to the following fields across all sockets: so_qlen so_incqlen so_qstate so_comp so_incomp so_list so_head While providing only coarse granularity, this approach avoids lock order issues between sockets by avoiding ownership of the fields by a specific socket and its per-socket mutexes. While here, rewrite soclose(), sofree(), soaccept(), and sonewconn() to add assertions, close additional races and address lock order concerns. In particular: - Reorganize the optimistic concurrency behavior in accept1() to always allocate a file descriptor with falloc() so that if we do find a socket, we don't have to encounter the "Oh, there wasn't a socket" race that can occur if falloc() sleeps in the current code, which broke inbound accept() ordering, not to mention requiring backing out socket state changes in a way that raced with the protocol level. We may want to add a lockless read of the queue state if polling of empty queues proves to be important to optimize. - In accept1(), soref() the socket while holding the accept lock so that the socket cannot be free'd in a race with the protocol layer. Likewise in netgraph equivilents of the accept1() code. - In sonewconn(), loop waiting for the queue to be small enough to insert our new socket once we've committed to inserting it, or races can occur that cause the incomplete socket queue to overfill. In the previously implementation, it was sufficient to simply tested once since calling soabort() didn't release synchronization permitting another thread to insert a socket as we discard a previous one. - In soclose()/sofree()/et al, it is the responsibility of the caller to remove a socket from the incomplete connection queue before calling soabort(), which prevents soabort() from having to walk into the accept socket to release the socket from its queue, and avoids races when releasing the accept mutex to enter soabort(), permitting soabort() to avoid lock ordering issues with the caller. - Generally cluster accept queue related operations together throughout these functions in order to facilitate locking. Annotate new locking in socketvar.h.
2004-06-02 04:15:39 +00:00
}
while (TAILQ_EMPTY(&head->so_comp) && head->so_error == 0) {
if (head->so_rcv.sb_state & SBS_CANTRCVMORE) {
head->so_error = ECONNABORTED;
1994-05-24 10:09:53 +00:00
break;
}
Integrate accept locking from rwatson_netperf, introducing a new global mutex, accept_mtx, which serializes access to the following fields across all sockets: so_qlen so_incqlen so_qstate so_comp so_incomp so_list so_head While providing only coarse granularity, this approach avoids lock order issues between sockets by avoiding ownership of the fields by a specific socket and its per-socket mutexes. While here, rewrite soclose(), sofree(), soaccept(), and sonewconn() to add assertions, close additional races and address lock order concerns. In particular: - Reorganize the optimistic concurrency behavior in accept1() to always allocate a file descriptor with falloc() so that if we do find a socket, we don't have to encounter the "Oh, there wasn't a socket" race that can occur if falloc() sleeps in the current code, which broke inbound accept() ordering, not to mention requiring backing out socket state changes in a way that raced with the protocol level. We may want to add a lockless read of the queue state if polling of empty queues proves to be important to optimize. - In accept1(), soref() the socket while holding the accept lock so that the socket cannot be free'd in a race with the protocol layer. Likewise in netgraph equivilents of the accept1() code. - In sonewconn(), loop waiting for the queue to be small enough to insert our new socket once we've committed to inserting it, or races can occur that cause the incomplete socket queue to overfill. In the previously implementation, it was sufficient to simply tested once since calling soabort() didn't release synchronization permitting another thread to insert a socket as we discard a previous one. - In soclose()/sofree()/et al, it is the responsibility of the caller to remove a socket from the incomplete connection queue before calling soabort(), which prevents soabort() from having to walk into the accept socket to release the socket from its queue, and avoids races when releasing the accept mutex to enter soabort(), permitting soabort() to avoid lock ordering issues with the caller. - Generally cluster accept queue related operations together throughout these functions in order to facilitate locking. Annotate new locking in socketvar.h.
2004-06-02 04:15:39 +00:00
error = msleep(&head->so_timeo, &accept_mtx, PSOCK | PCATCH,
"accept", 0);
if (error != 0) {
Integrate accept locking from rwatson_netperf, introducing a new global mutex, accept_mtx, which serializes access to the following fields across all sockets: so_qlen so_incqlen so_qstate so_comp so_incomp so_list so_head While providing only coarse granularity, this approach avoids lock order issues between sockets by avoiding ownership of the fields by a specific socket and its per-socket mutexes. While here, rewrite soclose(), sofree(), soaccept(), and sonewconn() to add assertions, close additional races and address lock order concerns. In particular: - Reorganize the optimistic concurrency behavior in accept1() to always allocate a file descriptor with falloc() so that if we do find a socket, we don't have to encounter the "Oh, there wasn't a socket" race that can occur if falloc() sleeps in the current code, which broke inbound accept() ordering, not to mention requiring backing out socket state changes in a way that raced with the protocol level. We may want to add a lockless read of the queue state if polling of empty queues proves to be important to optimize. - In accept1(), soref() the socket while holding the accept lock so that the socket cannot be free'd in a race with the protocol layer. Likewise in netgraph equivilents of the accept1() code. - In sonewconn(), loop waiting for the queue to be small enough to insert our new socket once we've committed to inserting it, or races can occur that cause the incomplete socket queue to overfill. In the previously implementation, it was sufficient to simply tested once since calling soabort() didn't release synchronization permitting another thread to insert a socket as we discard a previous one. - In soclose()/sofree()/et al, it is the responsibility of the caller to remove a socket from the incomplete connection queue before calling soabort(), which prevents soabort() from having to walk into the accept socket to release the socket from its queue, and avoids races when releasing the accept mutex to enter soabort(), permitting soabort() to avoid lock ordering issues with the caller. - Generally cluster accept queue related operations together throughout these functions in order to facilitate locking. Annotate new locking in socketvar.h.
2004-06-02 04:15:39 +00:00
ACCEPT_UNLOCK();
goto noconnection;
1994-05-24 10:09:53 +00:00
}
}
if (head->so_error) {
error = head->so_error;
head->so_error = 0;
Integrate accept locking from rwatson_netperf, introducing a new global mutex, accept_mtx, which serializes access to the following fields across all sockets: so_qlen so_incqlen so_qstate so_comp so_incomp so_list so_head While providing only coarse granularity, this approach avoids lock order issues between sockets by avoiding ownership of the fields by a specific socket and its per-socket mutexes. While here, rewrite soclose(), sofree(), soaccept(), and sonewconn() to add assertions, close additional races and address lock order concerns. In particular: - Reorganize the optimistic concurrency behavior in accept1() to always allocate a file descriptor with falloc() so that if we do find a socket, we don't have to encounter the "Oh, there wasn't a socket" race that can occur if falloc() sleeps in the current code, which broke inbound accept() ordering, not to mention requiring backing out socket state changes in a way that raced with the protocol level. We may want to add a lockless read of the queue state if polling of empty queues proves to be important to optimize. - In accept1(), soref() the socket while holding the accept lock so that the socket cannot be free'd in a race with the protocol layer. Likewise in netgraph equivilents of the accept1() code. - In sonewconn(), loop waiting for the queue to be small enough to insert our new socket once we've committed to inserting it, or races can occur that cause the incomplete socket queue to overfill. In the previously implementation, it was sufficient to simply tested once since calling soabort() didn't release synchronization permitting another thread to insert a socket as we discard a previous one. - In soclose()/sofree()/et al, it is the responsibility of the caller to remove a socket from the incomplete connection queue before calling soabort(), which prevents soabort() from having to walk into the accept socket to release the socket from its queue, and avoids races when releasing the accept mutex to enter soabort(), permitting soabort() to avoid lock ordering issues with the caller. - Generally cluster accept queue related operations together throughout these functions in order to facilitate locking. Annotate new locking in socketvar.h.
2004-06-02 04:15:39 +00:00
ACCEPT_UNLOCK();
goto noconnection;
1994-05-24 10:09:53 +00:00
}
so = TAILQ_FIRST(&head->so_comp);
Integrate accept locking from rwatson_netperf, introducing a new global mutex, accept_mtx, which serializes access to the following fields across all sockets: so_qlen so_incqlen so_qstate so_comp so_incomp so_list so_head While providing only coarse granularity, this approach avoids lock order issues between sockets by avoiding ownership of the fields by a specific socket and its per-socket mutexes. While here, rewrite soclose(), sofree(), soaccept(), and sonewconn() to add assertions, close additional races and address lock order concerns. In particular: - Reorganize the optimistic concurrency behavior in accept1() to always allocate a file descriptor with falloc() so that if we do find a socket, we don't have to encounter the "Oh, there wasn't a socket" race that can occur if falloc() sleeps in the current code, which broke inbound accept() ordering, not to mention requiring backing out socket state changes in a way that raced with the protocol level. We may want to add a lockless read of the queue state if polling of empty queues proves to be important to optimize. - In accept1(), soref() the socket while holding the accept lock so that the socket cannot be free'd in a race with the protocol layer. Likewise in netgraph equivilents of the accept1() code. - In sonewconn(), loop waiting for the queue to be small enough to insert our new socket once we've committed to inserting it, or races can occur that cause the incomplete socket queue to overfill. In the previously implementation, it was sufficient to simply tested once since calling soabort() didn't release synchronization permitting another thread to insert a socket as we discard a previous one. - In soclose()/sofree()/et al, it is the responsibility of the caller to remove a socket from the incomplete connection queue before calling soabort(), which prevents soabort() from having to walk into the accept socket to release the socket from its queue, and avoids races when releasing the accept mutex to enter soabort(), permitting soabort() to avoid lock ordering issues with the caller. - Generally cluster accept queue related operations together throughout these functions in order to facilitate locking. Annotate new locking in socketvar.h.
2004-06-02 04:15:39 +00:00
KASSERT(!(so->so_qstate & SQ_INCOMP), ("accept1: so SQ_INCOMP"));
KASSERT(so->so_qstate & SQ_COMP, ("accept1: so not SQ_COMP"));
/*
* Before changing the flags on the socket, we have to bump the
* reference count. Otherwise, if the protocol calls sofree(),
* the socket will be released due to a zero refcount.
*/
SOCK_LOCK(so); /* soref() and so_state update */
Integrate accept locking from rwatson_netperf, introducing a new global mutex, accept_mtx, which serializes access to the following fields across all sockets: so_qlen so_incqlen so_qstate so_comp so_incomp so_list so_head While providing only coarse granularity, this approach avoids lock order issues between sockets by avoiding ownership of the fields by a specific socket and its per-socket mutexes. While here, rewrite soclose(), sofree(), soaccept(), and sonewconn() to add assertions, close additional races and address lock order concerns. In particular: - Reorganize the optimistic concurrency behavior in accept1() to always allocate a file descriptor with falloc() so that if we do find a socket, we don't have to encounter the "Oh, there wasn't a socket" race that can occur if falloc() sleeps in the current code, which broke inbound accept() ordering, not to mention requiring backing out socket state changes in a way that raced with the protocol level. We may want to add a lockless read of the queue state if polling of empty queues proves to be important to optimize. - In accept1(), soref() the socket while holding the accept lock so that the socket cannot be free'd in a race with the protocol layer. Likewise in netgraph equivilents of the accept1() code. - In sonewconn(), loop waiting for the queue to be small enough to insert our new socket once we've committed to inserting it, or races can occur that cause the incomplete socket queue to overfill. In the previously implementation, it was sufficient to simply tested once since calling soabort() didn't release synchronization permitting another thread to insert a socket as we discard a previous one. - In soclose()/sofree()/et al, it is the responsibility of the caller to remove a socket from the incomplete connection queue before calling soabort(), which prevents soabort() from having to walk into the accept socket to release the socket from its queue, and avoids races when releasing the accept mutex to enter soabort(), permitting soabort() to avoid lock ordering issues with the caller. - Generally cluster accept queue related operations together throughout these functions in order to facilitate locking. Annotate new locking in socketvar.h.
2004-06-02 04:15:39 +00:00
soref(so); /* file descriptor reference */
TAILQ_REMOVE(&head->so_comp, so, so_list);
head->so_qlen--;
if (flags & ACCEPT4_INHERIT)
so->so_state |= (head->so_state & SS_NBIO);
else
so->so_state |= (flags & SOCK_NONBLOCK) ? SS_NBIO : 0;
Integrate accept locking from rwatson_netperf, introducing a new global mutex, accept_mtx, which serializes access to the following fields across all sockets: so_qlen so_incqlen so_qstate so_comp so_incomp so_list so_head While providing only coarse granularity, this approach avoids lock order issues between sockets by avoiding ownership of the fields by a specific socket and its per-socket mutexes. While here, rewrite soclose(), sofree(), soaccept(), and sonewconn() to add assertions, close additional races and address lock order concerns. In particular: - Reorganize the optimistic concurrency behavior in accept1() to always allocate a file descriptor with falloc() so that if we do find a socket, we don't have to encounter the "Oh, there wasn't a socket" race that can occur if falloc() sleeps in the current code, which broke inbound accept() ordering, not to mention requiring backing out socket state changes in a way that raced with the protocol level. We may want to add a lockless read of the queue state if polling of empty queues proves to be important to optimize. - In accept1(), soref() the socket while holding the accept lock so that the socket cannot be free'd in a race with the protocol layer. Likewise in netgraph equivilents of the accept1() code. - In sonewconn(), loop waiting for the queue to be small enough to insert our new socket once we've committed to inserting it, or races can occur that cause the incomplete socket queue to overfill. In the previously implementation, it was sufficient to simply tested once since calling soabort() didn't release synchronization permitting another thread to insert a socket as we discard a previous one. - In soclose()/sofree()/et al, it is the responsibility of the caller to remove a socket from the incomplete connection queue before calling soabort(), which prevents soabort() from having to walk into the accept socket to release the socket from its queue, and avoids races when releasing the accept mutex to enter soabort(), permitting soabort() to avoid lock ordering issues with the caller. - Generally cluster accept queue related operations together throughout these functions in order to facilitate locking. Annotate new locking in socketvar.h.
2004-06-02 04:15:39 +00:00
so->so_qstate &= ~SQ_COMP;
so->so_head = NULL;
SOCK_UNLOCK(so);
Integrate accept locking from rwatson_netperf, introducing a new global mutex, accept_mtx, which serializes access to the following fields across all sockets: so_qlen so_incqlen so_qstate so_comp so_incomp so_list so_head While providing only coarse granularity, this approach avoids lock order issues between sockets by avoiding ownership of the fields by a specific socket and its per-socket mutexes. While here, rewrite soclose(), sofree(), soaccept(), and sonewconn() to add assertions, close additional races and address lock order concerns. In particular: - Reorganize the optimistic concurrency behavior in accept1() to always allocate a file descriptor with falloc() so that if we do find a socket, we don't have to encounter the "Oh, there wasn't a socket" race that can occur if falloc() sleeps in the current code, which broke inbound accept() ordering, not to mention requiring backing out socket state changes in a way that raced with the protocol level. We may want to add a lockless read of the queue state if polling of empty queues proves to be important to optimize. - In accept1(), soref() the socket while holding the accept lock so that the socket cannot be free'd in a race with the protocol layer. Likewise in netgraph equivilents of the accept1() code. - In sonewconn(), loop waiting for the queue to be small enough to insert our new socket once we've committed to inserting it, or races can occur that cause the incomplete socket queue to overfill. In the previously implementation, it was sufficient to simply tested once since calling soabort() didn't release synchronization permitting another thread to insert a socket as we discard a previous one. - In soclose()/sofree()/et al, it is the responsibility of the caller to remove a socket from the incomplete connection queue before calling soabort(), which prevents soabort() from having to walk into the accept socket to release the socket from its queue, and avoids races when releasing the accept mutex to enter soabort(), permitting soabort() to avoid lock ordering issues with the caller. - Generally cluster accept queue related operations together throughout these functions in order to facilitate locking. Annotate new locking in socketvar.h.
2004-06-02 04:15:39 +00:00
ACCEPT_UNLOCK();
/* An extra reference on `nfp' has been held for us by falloc(). */
td->td_retval[0] = fd;
/* connection has been removed from the listen queue */
KNOTE_UNLOCKED(&head->so_rcv.sb_sel.si_note, 0);
if (flags & ACCEPT4_INHERIT) {
pgid = fgetown(&head->so_sigio);
if (pgid != 0)
fsetown(pgid, &so->so_sigio);
} else {
fflag &= ~(FNONBLOCK | FASYNC);
if (flags & SOCK_NONBLOCK)
fflag |= FNONBLOCK;
}
finit(nfp, fflag, DTYPE_SOCKET, so, &socketops);
/* Sync socket nonblocking/async state with file flags */
tmp = fflag & FNONBLOCK;
(void) fo_ioctl(nfp, FIONBIO, &tmp, td->td_ucred, td);
tmp = fflag & FASYNC;
(void) fo_ioctl(nfp, FIOASYNC, &tmp, td->td_ucred, td);
sa = NULL;
error = soaccept(so, &sa);
if (error != 0)
goto noconnection;
if (sa == NULL) {
if (name)
*namelen = 0;
goto done;
}
AUDIT_ARG_SOCKADDR(td, AT_FDCWD, sa);
if (name) {
/* check sa_len before it is destroyed */
if (*namelen > sa->sa_len)
*namelen = sa->sa_len;
#ifdef KTRACE
if (KTRPOINT(td, KTR_STRUCT))
ktrsockaddr(sa);
#endif
*name = sa;
sa = NULL;
1994-05-24 10:09:53 +00:00
}
noconnection:
free(sa, M_SONAME);
/*
* close the new descriptor, assuming someone hasn't ripped it
* out from under us.
*/
if (error != 0)
fdclose(td, nfp, fd);
/*
* Release explicitly held references before returning. We return
* a reference on nfp to the caller on success if they request it.
*/
done:
if (fp != NULL) {
if (error == 0) {
*fp = nfp;
nfp = NULL;
} else
*fp = NULL;
}
if (nfp != NULL)
fdrop(nfp, td);
fdrop(headfp, td);
1994-05-24 10:09:53 +00:00
return (error);
}
int
sys_accept(td, uap)
struct thread *td;
struct accept_args *uap;
{
return (accept1(td, uap->s, uap->name, uap->anamelen, ACCEPT4_INHERIT));
}
int
sys_accept4(td, uap)
struct thread *td;
struct accept4_args *uap;
{
if (uap->flags & ~(SOCK_CLOEXEC | SOCK_NONBLOCK))
return (EINVAL);
return (accept1(td, uap->s, uap->name, uap->anamelen, uap->flags));
}
#ifdef COMPAT_OLDSOCK
int
oaccept(td, uap)
struct thread *td;
struct accept_args *uap;
{
return (accept1(td, uap->s, uap->name, uap->anamelen,
ACCEPT4_INHERIT | ACCEPT4_COMPAT));
}
#endif /* COMPAT_OLDSOCK */
1994-05-24 10:09:53 +00:00
/* ARGSUSED */
int
sys_connect(td, uap)
struct thread *td;
struct connect_args /* {
int s;
caddr_t name;
int namelen;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct sockaddr *sa;
int error;
error = getsockaddr(&sa, uap->name, uap->namelen);
if (error == 0) {
error = kern_connectat(td, AT_FDCWD, uap->s, sa);
free(sa, M_SONAME);
}
return (error);
}
int
kern_connectat(struct thread *td, int dirfd, int fd, struct sockaddr *sa)
{
struct socket *so;
struct file *fp;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
int error, interrupted = 0;
1994-05-24 10:09:53 +00:00
AUDIT_ARG_FD(fd);
AUDIT_ARG_SOCKADDR(td, dirfd, sa);
error = getsock_cap(td, fd, cap_rights_init(&rights, CAP_CONNECT),
&fp, NULL);
if (error != 0)
return (error);
so = fp->f_data;
if (so->so_state & SS_ISCONNECTING) {
error = EALREADY;
goto done1;
}
#ifdef KTRACE
if (KTRPOINT(td, KTR_STRUCT))
ktrsockaddr(sa);
#endif
#ifdef MAC
error = mac_socket_check_connect(td->td_ucred, so, sa);
if (error != 0)
goto bad;
#endif
if (dirfd == AT_FDCWD)
error = soconnect(so, sa, td);
else
error = soconnectat(dirfd, so, sa, td);
if (error != 0)
1994-05-24 10:09:53 +00:00
goto bad;
if ((so->so_state & SS_NBIO) && (so->so_state & SS_ISCONNECTING)) {
error = EINPROGRESS;
goto done1;
1994-05-24 10:09:53 +00:00
}
SOCK_LOCK(so);
while ((so->so_state & SS_ISCONNECTING) && so->so_error == 0) {
error = msleep(&so->so_timeo, SOCK_MTX(so), PSOCK | PCATCH,
"connec", 0);
if (error != 0) {
if (error == EINTR || error == ERESTART)
interrupted = 1;
1994-05-24 10:09:53 +00:00
break;
}
}
1994-05-24 10:09:53 +00:00
if (error == 0) {
error = so->so_error;
so->so_error = 0;
}
SOCK_UNLOCK(so);
1994-05-24 10:09:53 +00:00
bad:
if (!interrupted)
so->so_state &= ~SS_ISCONNECTING;
1994-05-24 10:09:53 +00:00
if (error == ERESTART)
error = EINTR;
done1:
fdrop(fp, td);
1994-05-24 10:09:53 +00:00
return (error);
}
/* ARGSUSED */
int
sys_connectat(td, uap)
struct thread *td;
struct connectat_args /* {
int fd;
int s;
caddr_t name;
int namelen;
} */ *uap;
{
struct sockaddr *sa;
int error;
error = getsockaddr(&sa, uap->name, uap->namelen);
if (error == 0) {
error = kern_connectat(td, uap->fd, uap->s, sa);
free(sa, M_SONAME);
}
return (error);
}
int
kern_socketpair(struct thread *td, int domain, int type, int protocol,
int *rsv)
1994-05-24 10:09:53 +00:00
{
struct file *fp1, *fp2;
struct socket *so1, *so2;
int fd, error, oflag, fflag;
1994-05-24 10:09:53 +00:00
AUDIT_ARG_SOCKET(domain, type, protocol);
oflag = 0;
fflag = 0;
if ((type & SOCK_CLOEXEC) != 0) {
type &= ~SOCK_CLOEXEC;
oflag |= O_CLOEXEC;
}
if ((type & SOCK_NONBLOCK) != 0) {
type &= ~SOCK_NONBLOCK;
fflag |= FNONBLOCK;
}
#ifdef MAC
/* We might want to have a separate check for socket pairs. */
error = mac_socket_check_create(td->td_ucred, domain, type,
protocol);
if (error != 0)
return (error);
#endif
error = socreate(domain, &so1, type, protocol, td->td_ucred, td);
if (error != 0)
return (error);
error = socreate(domain, &so2, type, protocol, td->td_ucred, td);
if (error != 0)
1994-05-24 10:09:53 +00:00
goto free1;
/* On success extra reference to `fp1' and 'fp2' is set by falloc. */
error = falloc(td, &fp1, &fd, oflag);
if (error != 0)
1994-05-24 10:09:53 +00:00
goto free2;
rsv[0] = fd;
fp1->f_data = so1; /* so1 already has ref count */
error = falloc(td, &fp2, &fd, oflag);
if (error != 0)
1994-05-24 10:09:53 +00:00
goto free3;
fp2->f_data = so2; /* so2 already has ref count */
rsv[1] = fd;
error = soconnect2(so1, so2);
if (error != 0)
1994-05-24 10:09:53 +00:00
goto free4;
if (type == SOCK_DGRAM) {
1994-05-24 10:09:53 +00:00
/*
* Datagram socket connection is asymmetric.
*/
error = soconnect2(so2, so1);
if (error != 0)
1994-05-24 10:09:53 +00:00
goto free4;
}
finit(fp1, FREAD | FWRITE | fflag, DTYPE_SOCKET, fp1->f_data,
&socketops);
finit(fp2, FREAD | FWRITE | fflag, DTYPE_SOCKET, fp2->f_data,
&socketops);
if ((fflag & FNONBLOCK) != 0) {
(void) fo_ioctl(fp1, FIONBIO, &fflag, td->td_ucred, td);
(void) fo_ioctl(fp2, FIONBIO, &fflag, td->td_ucred, td);
}
fdrop(fp1, td);
fdrop(fp2, td);
return (0);
1994-05-24 10:09:53 +00:00
free4:
fdclose(td, fp2, rsv[1]);
fdrop(fp2, td);
1994-05-24 10:09:53 +00:00
free3:
fdclose(td, fp1, rsv[0]);
fdrop(fp1, td);
1994-05-24 10:09:53 +00:00
free2:
if (so2 != NULL)
(void)soclose(so2);
1994-05-24 10:09:53 +00:00
free1:
if (so1 != NULL)
(void)soclose(so1);
1994-05-24 10:09:53 +00:00
return (error);
}
int
sys_socketpair(struct thread *td, struct socketpair_args *uap)
{
int error, sv[2];
error = kern_socketpair(td, uap->domain, uap->type,
uap->protocol, sv);
if (error != 0)
return (error);
error = copyout(sv, uap->rsv, 2 * sizeof(int));
if (error != 0) {
(void)kern_close(td, sv[0]);
(void)kern_close(td, sv[1]);
}
return (error);
}
1998-02-09 06:11:36 +00:00
static int
sendit(td, s, mp, flags)
struct thread *td;
int s;
struct msghdr *mp;
int flags;
{
struct mbuf *control;
struct sockaddr *to;
int error;
#ifdef CAPABILITY_MODE
if (IN_CAPABILITY_MODE(td) && (mp->msg_name != NULL))
return (ECAPMODE);
#endif
if (mp->msg_name != NULL) {
error = getsockaddr(&to, mp->msg_name, mp->msg_namelen);
if (error != 0) {
to = NULL;
goto bad;
}
mp->msg_name = to;
} else {
to = NULL;
}
if (mp->msg_control) {
if (mp->msg_controllen < sizeof(struct cmsghdr)
#ifdef COMPAT_OLDSOCK
&& mp->msg_flags != MSG_COMPAT
#endif
) {
error = EINVAL;
goto bad;
}
error = sockargs(&control, mp->msg_control,
mp->msg_controllen, MT_CONTROL);
if (error != 0)
goto bad;
#ifdef COMPAT_OLDSOCK
if (mp->msg_flags == MSG_COMPAT) {
struct cmsghdr *cm;
M_PREPEND(control, sizeof(*cm), M_WAITOK);
cm = mtod(control, struct cmsghdr *);
cm->cmsg_len = control->m_len;
cm->cmsg_level = SOL_SOCKET;
cm->cmsg_type = SCM_RIGHTS;
}
#endif
} else {
control = NULL;
}
error = kern_sendit(td, s, mp, flags, control, UIO_USERSPACE);
bad:
free(to, M_SONAME);
return (error);
}
int
kern_sendit(td, s, mp, flags, control, segflg)
struct thread *td;
int s;
struct msghdr *mp;
int flags;
struct mbuf *control;
enum uio_seg segflg;
{
struct file *fp;
struct uio auio;
struct iovec *iov;
struct socket *so;
cap_rights_t rights;
#ifdef KTRACE
struct uio *ktruio = NULL;
#endif
ssize_t len;
int i, error;
AUDIT_ARG_FD(s);
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_init(&rights, CAP_SEND);
if (mp->msg_name != NULL) {
AUDIT_ARG_SOCKADDR(td, AT_FDCWD, mp->msg_name);
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_set(&rights, CAP_CONNECT);
}
error = getsock_cap(td, s, &rights, &fp, NULL);
if (error != 0)
return (error);
so = (struct socket *)fp->f_data;
#ifdef KTRACE
if (mp->msg_name != NULL && KTRPOINT(td, KTR_STRUCT))
ktrsockaddr(mp->msg_name);
#endif
#ifdef MAC
if (mp->msg_name != NULL) {
error = mac_socket_check_connect(td->td_ucred, so,
mp->msg_name);
if (error != 0)
goto bad;
}
error = mac_socket_check_send(td->td_ucred, so);
if (error != 0)
goto bad;
#endif
auio.uio_iov = mp->msg_iov;
auio.uio_iovcnt = mp->msg_iovlen;
auio.uio_segflg = segflg;
auio.uio_rw = UIO_WRITE;
auio.uio_td = td;
auio.uio_offset = 0; /* XXX */
auio.uio_resid = 0;
iov = mp->msg_iov;
for (i = 0; i < mp->msg_iovlen; i++, iov++) {
if ((auio.uio_resid += iov->iov_len) < 0) {
error = EINVAL;
goto bad;
}
}
#ifdef KTRACE
if (KTRPOINT(td, KTR_GENIO))
ktruio = cloneuio(&auio);
#endif
len = auio.uio_resid;
error = sosend(so, mp->msg_name, &auio, 0, control, flags, td);
if (error != 0) {
if (auio.uio_resid != len && (error == ERESTART ||
error == EINTR || error == EWOULDBLOCK))
error = 0;
/* Generation of SIGPIPE can be controlled per socket */
if (error == EPIPE && !(so->so_options & SO_NOSIGPIPE) &&
!(flags & MSG_NOSIGNAL)) {
PROC_LOCK(td->td_proc);
tdsignal(td, SIGPIPE);
PROC_UNLOCK(td->td_proc);
}
}
if (error == 0)
td->td_retval[0] = len - auio.uio_resid;
#ifdef KTRACE
if (ktruio != NULL) {
ktruio->uio_resid = td->td_retval[0];
ktrgenio(s, UIO_WRITE, ktruio, error);
}
#endif
bad:
fdrop(fp, td);
return (error);
}
int
sys_sendto(td, uap)
struct thread *td;
struct sendto_args /* {
int s;
caddr_t buf;
size_t len;
int flags;
caddr_t to;
int tolen;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct msghdr msg;
struct iovec aiov;
msg.msg_name = uap->to;
msg.msg_namelen = uap->tolen;
msg.msg_iov = &aiov;
msg.msg_iovlen = 1;
msg.msg_control = 0;
#ifdef COMPAT_OLDSOCK
msg.msg_flags = 0;
#endif
aiov.iov_base = uap->buf;
aiov.iov_len = uap->len;
return (sendit(td, uap->s, &msg, uap->flags));
1994-05-24 10:09:53 +00:00
}
#ifdef COMPAT_OLDSOCK
int
osend(td, uap)
struct thread *td;
struct osend_args /* {
int s;
caddr_t buf;
int len;
int flags;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct msghdr msg;
struct iovec aiov;
msg.msg_name = 0;
msg.msg_namelen = 0;
msg.msg_iov = &aiov;
msg.msg_iovlen = 1;
aiov.iov_base = uap->buf;
aiov.iov_len = uap->len;
msg.msg_control = 0;
msg.msg_flags = 0;
return (sendit(td, uap->s, &msg, uap->flags));
1994-05-24 10:09:53 +00:00
}
int
osendmsg(td, uap)
struct thread *td;
struct osendmsg_args /* {
int s;
caddr_t msg;
int flags;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct msghdr msg;
struct iovec *iov;
1994-05-24 10:09:53 +00:00
int error;
error = copyin(uap->msg, &msg, sizeof (struct omsghdr));
if (error != 0)
return (error);
error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
if (error != 0)
return (error);
1994-05-24 10:09:53 +00:00
msg.msg_iov = iov;
msg.msg_flags = MSG_COMPAT;
error = sendit(td, uap->s, &msg, uap->flags);
free(iov, M_IOV);
1994-05-24 10:09:53 +00:00
return (error);
}
#endif
int
sys_sendmsg(td, uap)
struct thread *td;
struct sendmsg_args /* {
int s;
caddr_t msg;
int flags;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct msghdr msg;
struct iovec *iov;
1994-05-24 10:09:53 +00:00
int error;
error = copyin(uap->msg, &msg, sizeof (msg));
if (error != 0)
return (error);
error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
if (error != 0)
return (error);
1994-05-24 10:09:53 +00:00
msg.msg_iov = iov;
#ifdef COMPAT_OLDSOCK
msg.msg_flags = 0;
#endif
error = sendit(td, uap->s, &msg, uap->flags);
free(iov, M_IOV);
1994-05-24 10:09:53 +00:00
return (error);
}
int
kern_recvit(td, s, mp, fromseg, controlp)
struct thread *td;
1994-05-24 10:09:53 +00:00
int s;
struct msghdr *mp;
enum uio_seg fromseg;
struct mbuf **controlp;
1994-05-24 10:09:53 +00:00
{
struct uio auio;
struct iovec *iov;
2013-02-07 00:27:11 +00:00
struct mbuf *m, *control = NULL;
caddr_t ctlbuf;
struct file *fp;
struct socket *so;
2013-02-07 00:27:11 +00:00
struct sockaddr *fromsa = NULL;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
1994-05-24 10:09:53 +00:00
#ifdef KTRACE
struct uio *ktruio = NULL;
1994-05-24 10:09:53 +00:00
#endif
ssize_t len;
int error, i;
1995-05-30 08:16:23 +00:00
if (controlp != NULL)
*controlp = NULL;
AUDIT_ARG_FD(s);
error = getsock_cap(td, s, cap_rights_init(&rights, CAP_RECV),
&fp, NULL);
if (error != 0)
1994-05-24 10:09:53 +00:00
return (error);
so = fp->f_data;
#ifdef MAC
error = mac_socket_check_receive(td->td_ucred, so);
if (error != 0) {
fdrop(fp, td);
return (error);
}
#endif
1994-05-24 10:09:53 +00:00
auio.uio_iov = mp->msg_iov;
auio.uio_iovcnt = mp->msg_iovlen;
auio.uio_segflg = UIO_USERSPACE;
auio.uio_rw = UIO_READ;
auio.uio_td = td;
1994-05-24 10:09:53 +00:00
auio.uio_offset = 0; /* XXX */
auio.uio_resid = 0;
iov = mp->msg_iov;
for (i = 0; i < mp->msg_iovlen; i++, iov++) {
if ((auio.uio_resid += iov->iov_len) < 0) {
fdrop(fp, td);
1994-05-24 10:09:53 +00:00
return (EINVAL);
}
1994-05-24 10:09:53 +00:00
}
#ifdef KTRACE
if (KTRPOINT(td, KTR_GENIO))
ktruio = cloneuio(&auio);
1994-05-24 10:09:53 +00:00
#endif
len = auio.uio_resid;
2013-02-07 00:27:11 +00:00
error = soreceive(so, &fromsa, &auio, NULL,
(mp->msg_control || controlp) ? &control : NULL,
&mp->msg_flags);
if (error != 0) {
if (auio.uio_resid != len && (error == ERESTART ||
1994-05-24 10:09:53 +00:00
error == EINTR || error == EWOULDBLOCK))
error = 0;
}
if (fromsa != NULL)
AUDIT_ARG_SOCKADDR(td, AT_FDCWD, fromsa);
1994-05-24 10:09:53 +00:00
#ifdef KTRACE
if (ktruio != NULL) {
ktruio->uio_resid = len - auio.uio_resid;
ktrgenio(s, UIO_READ, ktruio, error);
1994-05-24 10:09:53 +00:00
}
#endif
if (error != 0)
goto out;
td->td_retval[0] = len - auio.uio_resid;
if (mp->msg_name) {
len = mp->msg_namelen;
2013-02-07 00:27:11 +00:00
if (len <= 0 || fromsa == NULL)
len = 0;
else {
/* save sa_len before it is destroyed by MSG_COMPAT */
len = MIN(len, fromsa->sa_len);
1994-05-24 10:09:53 +00:00
#ifdef COMPAT_OLDSOCK
if (mp->msg_flags & MSG_COMPAT)
((struct osockaddr *)fromsa)->sa_family =
fromsa->sa_family;
#endif
if (fromseg == UIO_USERSPACE) {
error = copyout(fromsa, mp->msg_name,
(unsigned)len);
if (error != 0)
goto out;
} else
bcopy(fromsa, mp->msg_name, len);
}
mp->msg_namelen = len;
}
if (mp->msg_control && controlp == NULL) {
#ifdef COMPAT_OLDSOCK
/*
* We assume that old recvmsg calls won't receive access
* rights and other control info, esp. as control info
* is always optional and those options didn't exist in 4.3.
* If we receive rights, trim the cmsghdr; anything else
* is tossed.
*/
if (control && mp->msg_flags & MSG_COMPAT) {
if (mtod(control, struct cmsghdr *)->cmsg_level !=
SOL_SOCKET ||
mtod(control, struct cmsghdr *)->cmsg_type !=
SCM_RIGHTS) {
mp->msg_controllen = 0;
goto out;
}
control->m_len -= sizeof (struct cmsghdr);
control->m_data += sizeof (struct cmsghdr);
}
#endif
len = mp->msg_controllen;
m = control;
mp->msg_controllen = 0;
ctlbuf = mp->msg_control;
while (m && len > 0) {
unsigned int tocopy;
if (len >= m->m_len)
tocopy = m->m_len;
else {
mp->msg_flags |= MSG_CTRUNC;
tocopy = len;
}
if ((error = copyout(mtod(m, caddr_t),
ctlbuf, tocopy)) != 0)
goto out;
ctlbuf += tocopy;
len -= tocopy;
m = m->m_next;
}
mp->msg_controllen = ctlbuf - (caddr_t)mp->msg_control;
}
out:
fdrop(fp, td);
#ifdef KTRACE
if (fromsa && KTRPOINT(td, KTR_STRUCT))
ktrsockaddr(fromsa);
#endif
free(fromsa, M_SONAME);
2013-02-07 00:27:11 +00:00
if (error == 0 && controlp != NULL)
*controlp = control;
else if (control)
m_freem(control);
return (error);
}
static int
recvit(td, s, mp, namelenp)
struct thread *td;
int s;
struct msghdr *mp;
void *namelenp;
{
int error;
error = kern_recvit(td, s, mp, UIO_USERSPACE, NULL);
if (error != 0)
return (error);
if (namelenp != NULL) {
error = copyout(&mp->msg_namelen, namelenp, sizeof (socklen_t));
#ifdef COMPAT_OLDSOCK
if (mp->msg_flags & MSG_COMPAT)
error = 0; /* old recvfrom didn't check */
#endif
}
return (error);
}
int
sys_recvfrom(td, uap)
struct thread *td;
struct recvfrom_args /* {
int s;
caddr_t buf;
size_t len;
int flags;
struct sockaddr * __restrict from;
socklen_t * __restrict fromlenaddr;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct msghdr msg;
struct iovec aiov;
int error;
if (uap->fromlenaddr) {
error = copyin(uap->fromlenaddr,
&msg.msg_namelen, sizeof (msg.msg_namelen));
if (error != 0)
goto done2;
} else {
1994-05-24 10:09:53 +00:00
msg.msg_namelen = 0;
}
1994-05-24 10:09:53 +00:00
msg.msg_name = uap->from;
msg.msg_iov = &aiov;
msg.msg_iovlen = 1;
aiov.iov_base = uap->buf;
aiov.iov_len = uap->len;
msg.msg_control = 0;
msg.msg_flags = uap->flags;
2002-06-29 00:02:01 +00:00
error = recvit(td, uap->s, &msg, uap->fromlenaddr);
done2:
return (error);
1994-05-24 10:09:53 +00:00
}
#ifdef COMPAT_OLDSOCK
int
orecvfrom(td, uap)
struct thread *td;
struct recvfrom_args *uap;
{
uap->flags |= MSG_COMPAT;
return (sys_recvfrom(td, uap));
}
#endif
1994-05-24 10:09:53 +00:00
#ifdef COMPAT_OLDSOCK
int
orecv(td, uap)
struct thread *td;
struct orecv_args /* {
int s;
caddr_t buf;
int len;
int flags;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct msghdr msg;
struct iovec aiov;
msg.msg_name = 0;
msg.msg_namelen = 0;
msg.msg_iov = &aiov;
msg.msg_iovlen = 1;
aiov.iov_base = uap->buf;
aiov.iov_len = uap->len;
msg.msg_control = 0;
msg.msg_flags = uap->flags;
return (recvit(td, uap->s, &msg, NULL));
1994-05-24 10:09:53 +00:00
}
/*
* Old recvmsg. This code takes advantage of the fact that the old msghdr
* overlays the new one, missing only the flags, and with the (old) access
* rights where the control fields are now.
*/
int
orecvmsg(td, uap)
struct thread *td;
struct orecvmsg_args /* {
int s;
struct omsghdr *msg;
int flags;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct msghdr msg;
struct iovec *iov;
1994-05-24 10:09:53 +00:00
int error;
2002-06-29 00:02:01 +00:00
error = copyin(uap->msg, &msg, sizeof (struct omsghdr));
if (error != 0)
1994-05-24 10:09:53 +00:00
return (error);
error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
if (error != 0)
return (error);
msg.msg_flags = uap->flags | MSG_COMPAT;
1994-05-24 10:09:53 +00:00
msg.msg_iov = iov;
2002-06-29 00:02:01 +00:00
error = recvit(td, uap->s, &msg, &uap->msg->msg_namelen);
1994-05-24 10:09:53 +00:00
if (msg.msg_controllen && error == 0)
2002-06-29 00:02:01 +00:00
error = copyout(&msg.msg_controllen,
&uap->msg->msg_accrightslen, sizeof (int));
free(iov, M_IOV);
1994-05-24 10:09:53 +00:00
return (error);
}
#endif
int
sys_recvmsg(td, uap)
struct thread *td;
struct recvmsg_args /* {
int s;
struct msghdr *msg;
int flags;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct msghdr msg;
struct iovec *uiov, *iov;
int error;
1994-05-24 10:09:53 +00:00
2002-06-29 00:02:01 +00:00
error = copyin(uap->msg, &msg, sizeof (msg));
if (error != 0)
return (error);
error = copyiniov(msg.msg_iov, msg.msg_iovlen, &iov, EMSGSIZE);
if (error != 0)
return (error);
1994-05-24 10:09:53 +00:00
msg.msg_flags = uap->flags;
#ifdef COMPAT_OLDSOCK
msg.msg_flags &= ~MSG_COMPAT;
1994-05-24 10:09:53 +00:00
#endif
uiov = msg.msg_iov;
msg.msg_iov = iov;
2002-06-29 00:02:01 +00:00
error = recvit(td, uap->s, &msg, NULL);
if (error == 0) {
1994-05-24 10:09:53 +00:00
msg.msg_iov = uiov;
2002-06-29 00:02:01 +00:00
error = copyout(&msg, uap->msg, sizeof(msg));
1994-05-24 10:09:53 +00:00
}
free(iov, M_IOV);
1994-05-24 10:09:53 +00:00
return (error);
}
1994-05-24 10:09:53 +00:00
/* ARGSUSED */
int
sys_shutdown(td, uap)
struct thread *td;
struct shutdown_args /* {
int s;
int how;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
struct socket *so;
struct file *fp;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
1994-05-24 10:09:53 +00:00
int error;
AUDIT_ARG_FD(uap->s);
error = getsock_cap(td, uap->s, cap_rights_init(&rights, CAP_SHUTDOWN),
&fp, NULL);
if (error == 0) {
so = fp->f_data;
error = soshutdown(so, uap->how);
/*
* Previous versions did not return ENOTCONN, but 0 in
* case the socket was not connected. Some important
* programs like syslogd up to r279016, 2015-02-19,
* still depend on this behavior.
*/
if (error == ENOTCONN &&
td->td_proc->p_osrel < P_OSREL_SHUTDOWN_ENOTCONN)
error = 0;
fdrop(fp, td);
}
return (error);
1994-05-24 10:09:53 +00:00
}
/* ARGSUSED */
int
sys_setsockopt(td, uap)
struct thread *td;
struct setsockopt_args /* {
int s;
int level;
int name;
caddr_t val;
int valsize;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
return (kern_setsockopt(td, uap->s, uap->level, uap->name,
uap->val, UIO_USERSPACE, uap->valsize));
}
int
kern_setsockopt(td, s, level, name, val, valseg, valsize)
struct thread *td;
int s;
int level;
int name;
void *val;
enum uio_seg valseg;
socklen_t valsize;
{
struct socket *so;
struct file *fp;
struct sockopt sopt;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
int error;
1994-05-24 10:09:53 +00:00
if (val == NULL && valsize != 0)
return (EFAULT);
if ((int)valsize < 0)
return (EINVAL);
sopt.sopt_dir = SOPT_SET;
sopt.sopt_level = level;
sopt.sopt_name = name;
sopt.sopt_val = val;
sopt.sopt_valsize = valsize;
switch (valseg) {
case UIO_USERSPACE:
sopt.sopt_td = td;
break;
case UIO_SYSSPACE:
sopt.sopt_td = NULL;
break;
default:
panic("kern_setsockopt called with bad valseg");
}
AUDIT_ARG_FD(s);
error = getsock_cap(td, s, cap_rights_init(&rights, CAP_SETSOCKOPT),
&fp, NULL);
if (error == 0) {
so = fp->f_data;
error = sosetopt(so, &sopt);
fdrop(fp, td);
}
return(error);
1994-05-24 10:09:53 +00:00
}
/* ARGSUSED */
int
sys_getsockopt(td, uap)
struct thread *td;
struct getsockopt_args /* {
int s;
int level;
int name;
void * __restrict val;
socklen_t * __restrict avalsize;
} */ *uap;
1994-05-24 10:09:53 +00:00
{
socklen_t valsize;
int error;
1994-05-24 10:09:53 +00:00
if (uap->val) {
2002-06-29 00:02:01 +00:00
error = copyin(uap->avalsize, &valsize, sizeof (valsize));
if (error != 0)
return (error);
}
error = kern_getsockopt(td, uap->s, uap->level, uap->name,
uap->val, UIO_USERSPACE, &valsize);
if (error == 0)
2002-06-29 00:02:01 +00:00
error = copyout(&valsize, uap->avalsize, sizeof (valsize));
return (error);
}
/*
* Kernel version of getsockopt.
* optval can be a userland or userspace. optlen is always a kernel pointer.
*/
int
kern_getsockopt(td, s, level, name, val, valseg, valsize)
struct thread *td;
int s;
int level;
int name;
void *val;
enum uio_seg valseg;
socklen_t *valsize;
{
struct socket *so;
struct file *fp;
struct sockopt sopt;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
int error;
if (val == NULL)
*valsize = 0;
if ((int)*valsize < 0)
return (EINVAL);
sopt.sopt_dir = SOPT_GET;
sopt.sopt_level = level;
sopt.sopt_name = name;
sopt.sopt_val = val;
sopt.sopt_valsize = (size_t)*valsize; /* checked non-negative above */
switch (valseg) {
case UIO_USERSPACE:
sopt.sopt_td = td;
break;
case UIO_SYSSPACE:
sopt.sopt_td = NULL;
break;
default:
panic("kern_getsockopt called with bad valseg");
}
AUDIT_ARG_FD(s);
error = getsock_cap(td, s, cap_rights_init(&rights, CAP_GETSOCKOPT),
&fp, NULL);
if (error == 0) {
so = fp->f_data;
error = sogetopt(so, &sopt);
*valsize = sopt.sopt_valsize;
fdrop(fp, td);
1994-05-24 10:09:53 +00:00
}
return (error);
}
/*
* getsockname1() - Get socket name.
1994-05-24 10:09:53 +00:00
*/
/* ARGSUSED */
static int
getsockname1(td, uap, compat)
struct thread *td;
struct getsockname_args /* {
int fdes;
struct sockaddr * __restrict asa;
socklen_t * __restrict alen;
} */ *uap;
int compat;
1994-05-24 10:09:53 +00:00
{
struct sockaddr *sa;
socklen_t len;
int error;
error = copyin(uap->alen, &len, sizeof(len));
if (error != 0)
return (error);
error = kern_getsockname(td, uap->fdes, &sa, &len);
if (error != 0)
return (error);
if (len != 0) {
#ifdef COMPAT_OLDSOCK
if (compat)
((struct osockaddr *)sa)->sa_family = sa->sa_family;
#endif
error = copyout(sa, uap->asa, (u_int)len);
}
free(sa, M_SONAME);
if (error == 0)
error = copyout(&len, uap->alen, sizeof(len));
return (error);
}
int
kern_getsockname(struct thread *td, int fd, struct sockaddr **sa,
socklen_t *alen)
{
struct socket *so;
struct file *fp;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
socklen_t len;
int error;
1994-05-24 10:09:53 +00:00
AUDIT_ARG_FD(fd);
error = getsock_cap(td, fd, cap_rights_init(&rights, CAP_GETSOCKNAME),
&fp, NULL);
if (error != 0)
return (error);
so = fp->f_data;
*sa = NULL;
Change the curvnet variable from a global const struct vnet *, previously always pointing to the default vnet context, to a dynamically changing thread-local one. The currvnet context should be set on entry to networking code via CURVNET_SET() macros, and reverted to previous state via CURVNET_RESTORE(). Recursions on curvnet are permitted, though strongly discuouraged. This change should have no functional impact on nooptions VIMAGE kernel builds, where CURVNET_* macros expand to whitespace. The curthread->td_vnet (aka curvnet) variable's purpose is to be an indicator of the vnet context in which the current network-related operation takes place, in case we cannot deduce the current vnet context from any other source, such as by looking at mbuf's m->m_pkthdr.rcvif->if_vnet, sockets's so->so_vnet etc. Moreover, so far curvnet has turned out to be an invaluable consistency checking aid: it helps to catch cases when sockets, ifnets or any other vnet-aware structures may have leaked from one vnet to another. The exact placement of the CURVNET_SET() / CURVNET_RESTORE() macros was a result of an empirical iterative process, whith an aim to reduce recursions on CURVNET_SET() to a minimum, while still reducing the scope of CURVNET_SET() to networking only operations - the alternative would be calling CURVNET_SET() on each system call entry. In general, curvnet has to be set in three typicall cases: when processing socket-related requests from userspace or from within the kernel; when processing inbound traffic flowing from device drivers to upper layers of the networking stack, and when executing timer-driven networking functions. This change also introduces a DDB subcommand to show the list of all vnet instances. Approved by: julian (mentor)
2009-05-05 10:56:12 +00:00
CURVNET_SET(so->so_vnet);
error = (*so->so_proto->pr_usrreqs->pru_sockaddr)(so, sa);
Change the curvnet variable from a global const struct vnet *, previously always pointing to the default vnet context, to a dynamically changing thread-local one. The currvnet context should be set on entry to networking code via CURVNET_SET() macros, and reverted to previous state via CURVNET_RESTORE(). Recursions on curvnet are permitted, though strongly discuouraged. This change should have no functional impact on nooptions VIMAGE kernel builds, where CURVNET_* macros expand to whitespace. The curthread->td_vnet (aka curvnet) variable's purpose is to be an indicator of the vnet context in which the current network-related operation takes place, in case we cannot deduce the current vnet context from any other source, such as by looking at mbuf's m->m_pkthdr.rcvif->if_vnet, sockets's so->so_vnet etc. Moreover, so far curvnet has turned out to be an invaluable consistency checking aid: it helps to catch cases when sockets, ifnets or any other vnet-aware structures may have leaked from one vnet to another. The exact placement of the CURVNET_SET() / CURVNET_RESTORE() macros was a result of an empirical iterative process, whith an aim to reduce recursions on CURVNET_SET() to a minimum, while still reducing the scope of CURVNET_SET() to networking only operations - the alternative would be calling CURVNET_SET() on each system call entry. In general, curvnet has to be set in three typicall cases: when processing socket-related requests from userspace or from within the kernel; when processing inbound traffic flowing from device drivers to upper layers of the networking stack, and when executing timer-driven networking functions. This change also introduces a DDB subcommand to show the list of all vnet instances. Approved by: julian (mentor)
2009-05-05 10:56:12 +00:00
CURVNET_RESTORE();
if (error != 0)
1994-05-24 10:09:53 +00:00
goto bad;
if (*sa == NULL)
len = 0;
else
len = MIN(*alen, (*sa)->sa_len);
*alen = len;
#ifdef KTRACE
if (KTRPOINT(td, KTR_STRUCT))
ktrsockaddr(*sa);
#endif
1994-05-24 10:09:53 +00:00
bad:
fdrop(fp, td);
if (error != 0 && *sa != NULL) {
free(*sa, M_SONAME);
*sa = NULL;
}
1994-05-24 10:09:53 +00:00
return (error);
}
int
sys_getsockname(td, uap)
struct thread *td;
struct getsockname_args *uap;
1994-05-24 10:09:53 +00:00
{
return (getsockname1(td, uap, 0));
1994-05-24 10:09:53 +00:00
}
#ifdef COMPAT_OLDSOCK
int
ogetsockname(td, uap)
struct thread *td;
struct getsockname_args *uap;
1994-05-24 10:09:53 +00:00
{
return (getsockname1(td, uap, 1));
1994-05-24 10:09:53 +00:00
}
#endif /* COMPAT_OLDSOCK */
1994-05-24 10:09:53 +00:00
/*
* getpeername1() - Get name of peer for connected socket.
*/
/* ARGSUSED */
static int
getpeername1(td, uap, compat)
struct thread *td;
struct getpeername_args /* {
int fdes;
struct sockaddr * __restrict asa;
socklen_t * __restrict alen;
} */ *uap;
int compat;
1994-05-24 10:09:53 +00:00
{
struct sockaddr *sa;
socklen_t len;
int error;
error = copyin(uap->alen, &len, sizeof (len));
if (error != 0)
return (error);
error = kern_getpeername(td, uap->fdes, &sa, &len);
if (error != 0)
return (error);
if (len != 0) {
#ifdef COMPAT_OLDSOCK
if (compat)
((struct osockaddr *)sa)->sa_family = sa->sa_family;
#endif
error = copyout(sa, uap->asa, (u_int)len);
}
free(sa, M_SONAME);
if (error == 0)
error = copyout(&len, uap->alen, sizeof(len));
return (error);
}
int
kern_getpeername(struct thread *td, int fd, struct sockaddr **sa,
socklen_t *alen)
{
struct socket *so;
struct file *fp;
Change the cap_rights_t type from uint64_t to a structure that we can extend in the future in a backward compatible (API and ABI) way. The cap_rights_t represents capability rights. We used to use one bit to represent one right, but we are running out of spare bits. Currently the new structure provides place for 114 rights (so 50 more than the previous cap_rights_t), but it is possible to grow the structure to hold at least 285 rights, although we can make it even larger if 285 rights won't be enough. The structure definition looks like this: struct cap_rights { uint64_t cr_rights[CAP_RIGHTS_VERSION + 2]; }; The initial CAP_RIGHTS_VERSION is 0. The top two bits in the first element of the cr_rights[] array contain total number of elements in the array - 2. This means if those two bits are equal to 0, we have 2 array elements. The top two bits in all remaining array elements should be 0. The next five bits in all array elements contain array index. Only one bit is used and bit position in this five-bits range defines array index. This means there can be at most five array elements in the future. To define new right the CAPRIGHT() macro must be used. The macro takes two arguments - an array index and a bit to set, eg. #define CAP_PDKILL CAPRIGHT(1, 0x0000000000000800ULL) We still support aliases that combine few rights, but the rights have to belong to the same array element, eg: #define CAP_LOOKUP CAPRIGHT(0, 0x0000000000000400ULL) #define CAP_FCHMOD CAPRIGHT(0, 0x0000000000002000ULL) #define CAP_FCHMODAT (CAP_FCHMOD | CAP_LOOKUP) There is new API to manage the new cap_rights_t structure: cap_rights_t *cap_rights_init(cap_rights_t *rights, ...); void cap_rights_set(cap_rights_t *rights, ...); void cap_rights_clear(cap_rights_t *rights, ...); bool cap_rights_is_set(const cap_rights_t *rights, ...); bool cap_rights_is_valid(const cap_rights_t *rights); void cap_rights_merge(cap_rights_t *dst, const cap_rights_t *src); void cap_rights_remove(cap_rights_t *dst, const cap_rights_t *src); bool cap_rights_contains(const cap_rights_t *big, const cap_rights_t *little); Capability rights to the cap_rights_init(), cap_rights_set(), cap_rights_clear() and cap_rights_is_set() functions are provided by separating them with commas, eg: cap_rights_t rights; cap_rights_init(&rights, CAP_READ, CAP_WRITE, CAP_FSTAT); There is no need to terminate the list of rights, as those functions are actually macros that take care of the termination, eg: #define cap_rights_set(rights, ...) \ __cap_rights_set((rights), __VA_ARGS__, 0ULL) void __cap_rights_set(cap_rights_t *rights, ...); Thanks to using one bit as an array index we can assert in those functions that there are no two rights belonging to different array elements provided together. For example this is illegal and will be detected, because CAP_LOOKUP belongs to element 0 and CAP_PDKILL to element 1: cap_rights_init(&rights, CAP_LOOKUP | CAP_PDKILL); Providing several rights that belongs to the same array's element this way is correct, but is not advised. It should only be used for aliases definition. This commit also breaks compatibility with some existing Capsicum system calls, but I see no other way to do that. This should be fine as Capsicum is still experimental and this change is not going to 9.x. Sponsored by: The FreeBSD Foundation
2013-09-05 00:09:56 +00:00
cap_rights_t rights;
socklen_t len;
int error;
1994-05-24 10:09:53 +00:00
AUDIT_ARG_FD(fd);
error = getsock_cap(td, fd, cap_rights_init(&rights, CAP_GETPEERNAME),
&fp, NULL);
if (error != 0)
return (error);
so = fp->f_data;
if ((so->so_state & (SS_ISCONNECTED|SS_ISCONFIRMING)) == 0) {
error = ENOTCONN;
goto done;
}
*sa = NULL;
Change the curvnet variable from a global const struct vnet *, previously always pointing to the default vnet context, to a dynamically changing thread-local one. The currvnet context should be set on entry to networking code via CURVNET_SET() macros, and reverted to previous state via CURVNET_RESTORE(). Recursions on curvnet are permitted, though strongly discuouraged. This change should have no functional impact on nooptions VIMAGE kernel builds, where CURVNET_* macros expand to whitespace. The curthread->td_vnet (aka curvnet) variable's purpose is to be an indicator of the vnet context in which the current network-related operation takes place, in case we cannot deduce the current vnet context from any other source, such as by looking at mbuf's m->m_pkthdr.rcvif->if_vnet, sockets's so->so_vnet etc. Moreover, so far curvnet has turned out to be an invaluable consistency checking aid: it helps to catch cases when sockets, ifnets or any other vnet-aware structures may have leaked from one vnet to another. The exact placement of the CURVNET_SET() / CURVNET_RESTORE() macros was a result of an empirical iterative process, whith an aim to reduce recursions on CURVNET_SET() to a minimum, while still reducing the scope of CURVNET_SET() to networking only operations - the alternative would be calling CURVNET_SET() on each system call entry. In general, curvnet has to be set in three typicall cases: when processing socket-related requests from userspace or from within the kernel; when processing inbound traffic flowing from device drivers to upper layers of the networking stack, and when executing timer-driven networking functions. This change also introduces a DDB subcommand to show the list of all vnet instances. Approved by: julian (mentor)
2009-05-05 10:56:12 +00:00
CURVNET_SET(so->so_vnet);
error = (*so->so_proto->pr_usrreqs->pru_peeraddr)(so, sa);
Change the curvnet variable from a global const struct vnet *, previously always pointing to the default vnet context, to a dynamically changing thread-local one. The currvnet context should be set on entry to networking code via CURVNET_SET() macros, and reverted to previous state via CURVNET_RESTORE(). Recursions on curvnet are permitted, though strongly discuouraged. This change should have no functional impact on nooptions VIMAGE kernel builds, where CURVNET_* macros expand to whitespace. The curthread->td_vnet (aka curvnet) variable's purpose is to be an indicator of the vnet context in which the current network-related operation takes place, in case we cannot deduce the current vnet context from any other source, such as by looking at mbuf's m->m_pkthdr.rcvif->if_vnet, sockets's so->so_vnet etc. Moreover, so far curvnet has turned out to be an invaluable consistency checking aid: it helps to catch cases when sockets, ifnets or any other vnet-aware structures may have leaked from one vnet to another. The exact placement of the CURVNET_SET() / CURVNET_RESTORE() macros was a result of an empirical iterative process, whith an aim to reduce recursions on CURVNET_SET() to a minimum, while still reducing the scope of CURVNET_SET() to networking only operations - the alternative would be calling CURVNET_SET() on each system call entry. In general, curvnet has to be set in three typicall cases: when processing socket-related requests from userspace or from within the kernel; when processing inbound traffic flowing from device drivers to upper layers of the networking stack, and when executing timer-driven networking functions. This change also introduces a DDB subcommand to show the list of all vnet instances. Approved by: julian (mentor)
2009-05-05 10:56:12 +00:00
CURVNET_RESTORE();
if (error != 0)
1994-05-24 10:09:53 +00:00
goto bad;
if (*sa == NULL)
len = 0;
else
len = MIN(*alen, (*sa)->sa_len);
*alen = len;
#ifdef KTRACE
if (KTRPOINT(td, KTR_STRUCT))
ktrsockaddr(*sa);
#endif
1994-05-24 10:09:53 +00:00
bad:
if (error != 0 && *sa != NULL) {
free(*sa, M_SONAME);
*sa = NULL;
}
done:
fdrop(fp, td);
1994-05-24 10:09:53 +00:00
return (error);
}
int
sys_getpeername(td, uap)
struct thread *td;
struct getpeername_args *uap;
{
return (getpeername1(td, uap, 0));
}
#ifdef COMPAT_OLDSOCK
int
ogetpeername(td, uap)
struct thread *td;
struct ogetpeername_args *uap;
{
/* XXX uap should have type `getpeername_args *' to begin with. */
return (getpeername1(td, (struct getpeername_args *)uap, 1));
}
#endif /* COMPAT_OLDSOCK */
int
1994-05-24 10:09:53 +00:00
sockargs(mp, buf, buflen, type)
struct mbuf **mp;
caddr_t buf;
int buflen, type;
{
struct sockaddr *sa;
struct mbuf *m;
1994-05-24 10:09:53 +00:00
int error;
if (buflen > MLEN) {
1994-05-24 10:09:53 +00:00
#ifdef COMPAT_OLDSOCK
if (type == MT_SONAME && buflen <= 112)
1994-05-24 10:09:53 +00:00
buflen = MLEN; /* unix domain compat. hack */
else
#endif
if (buflen > MCLBYTES)
return (EINVAL);
1994-05-24 10:09:53 +00:00
}
m = m_get2(buflen, M_WAITOK, type, 0);
1994-05-24 10:09:53 +00:00
m->m_len = buflen;
error = copyin(buf, mtod(m, caddr_t), (u_int)buflen);
if (error != 0)
1994-05-24 10:09:53 +00:00
(void) m_free(m);
else {
*mp = m;
if (type == MT_SONAME) {
sa = mtod(m, struct sockaddr *);
#if defined(COMPAT_OLDSOCK) && BYTE_ORDER != BIG_ENDIAN
if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
sa->sa_family = sa->sa_len;
#endif
sa->sa_len = buflen;
}
}
return (error);
}
int
getsockaddr(namp, uaddr, len)
struct sockaddr **namp;
caddr_t uaddr;
size_t len;
{
struct sockaddr *sa;
int error;
if (len > SOCK_MAXADDRLEN)
2004-01-10 13:03:43 +00:00
return (ENAMETOOLONG);
if (len < offsetof(struct sockaddr, sa_data[0]))
2004-01-10 17:14:53 +00:00
return (EINVAL);
sa = malloc(len, M_SONAME, M_WAITOK);
error = copyin(uaddr, sa, len);
if (error != 0) {
free(sa, M_SONAME);
} else {
#if defined(COMPAT_OLDSOCK) && BYTE_ORDER != BIG_ENDIAN
if (sa->sa_family == 0 && sa->sa_len < AF_MAX)
sa->sa_family = sa->sa_len;
#endif
sa->sa_len = len;
*namp = sa;
}
2004-01-10 13:03:43 +00:00
return (error);
}