NetBSD/usr.sbin/perfused/perfused.c

491 lines
11 KiB
C
Raw Normal View History

2021-08-08 23:56:54 +03:00
/* $NetBSD: perfused.c,v 1.26 2021/08/08 20:56:54 nia Exp $ */
/*-
* Copyright (c) 2010 Emmanuel Dreyfus. 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.
*
* THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. 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 FOUNDATION 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.
*/
#include <stdio.h>
#include <unistd.h>
#include <stdlib.h>
#include <fcntl.h>
#include <syslog.h>
#include <ctype.h>
#include <paths.h>
#include <stdarg.h>
#include <err.h>
#include <errno.h>
#include <string.h>
#include <sysexits.h>
#include <signal.h>
#include <puffs.h>
#include <sys/wait.h>
#include <sys/param.h>
#include <sys/queue.h>
#include <sys/uio.h>
#include <sys/socket.h>
#include <sys/un.h>
#include <machine/vmparam.h>
#include "perfused.h"
/*
* This is used for trace file. of course it will not work if
* we ever mount multiple filesystems in a single perfused,
* but it is not sure we will ever want to do that.
*/
static struct puffs_usermount *perfused_mount = NULL;
static FILE *perfused_trace = NULL;
static int access_mount(const char *, uid_t, int);
static void new_mount(int, int);
static int parse_debug(char *);
static void siginfo_handler(int);
static void sigusr1_handler(int);
static int parse_options(int, char **);
static void get_mount_info(int, struct perfuse_mount_info *, int);
/*
* Flags for new_mount()
*/
#define PMNT_DEVFUSE 0x0 /* We use /dev/fuse */
#define PMNT_SOCKPAIR 0x1 /* We use socketpair */
#define PMNT_DGRAM 0x2 /* We use SOCK_DGRAM sockets */
static int
access_mount(const char *mnt, uid_t uid, int ro)
{
struct stat st;
mode_t mode;
if (uid == 0)
return 0;
if (stat(mnt, &st) == -1)
return -1;
if (st.st_uid != uid)
return -1;
mode = S_IRUSR;
if (!ro)
mode |= S_IWUSR;
if ((st.st_mode & mode) == mode)
return 0;
return -1;
}
static void
get_mount_info(int fd, struct perfuse_mount_info *pmi, int sock_type)
{
struct perfuse_mount_out *pmo;
struct sockcred cred;
int opt;
char *cp;
char *source = NULL;
char *target = NULL;
char *filesystemtype = NULL;
long mountflags = 0;
void *data = NULL;
char *sock = NULL;
pmo = (struct perfuse_mount_out *)
perfused_recv_early(fd, &cred, sizeof(cred));
if (pmo == NULL) {
if (shutdown(fd, SHUT_RDWR) != 0)
DERR(EX_OSERR, "shutdown failed");
exit(EX_PROTOCOL);
}
/*
* We do not need peer creds beyond this point
*/
opt = 0;
2021-08-08 23:56:54 +03:00
if (setsockopt(fd, SOL_LOCAL, LOCAL_CREDS, &opt, sizeof(opt)) != 0)
DWARN("%s: setsockopt LOCAL_CREDS failed", __func__);
#ifdef PERFUSE_DEBUG
if (perfuse_diagflags & PDF_MISC)
DPRINTF("perfuse lengths: source = %"PRId32", "
"target = %"PRId32", filesystemtype = %"PRId32", "
"data = %"PRId32", sock = %"PRId32"\n",
pmo->pmo_source_len, pmo->pmo_target_len,
pmo->pmo_filesystemtype_len, pmo->pmo_data_len,
pmo->pmo_sock_len);
#endif
cp = (char *)(void *)(pmo + 1);
if (pmo->pmo_source_len != 0) {
source = cp;
cp += pmo->pmo_source_len;
}
if (pmo->pmo_target_len != 0) {
target = cp;
cp += pmo->pmo_target_len;
}
if (pmo->pmo_filesystemtype_len != 0) {
filesystemtype = cp;
cp += pmo->pmo_filesystemtype_len;
}
mountflags = pmo->pmo_mountflags;
if (pmo->pmo_data_len != 0) {
data = cp;
cp += pmo->pmo_data_len;
}
if (pmo->pmo_sock_len != 0) {
sock = cp;
cp += pmo->pmo_sock_len;
}
#ifdef PERFUSE_DEBUG
if (perfuse_diagflags & PDF_MISC)
DPRINTF("%s(\"%s\", \"%s\", \"%s\", 0x%lx, \"%s\", \"%s\")\n",
__func__, source, target, filesystemtype,
mountflags, (const char *)data, sock);
#endif
pmi->pmi_source = source;
pmi->pmi_target = target;
pmi->pmi_filesystemtype = filesystemtype;
2010-09-06 05:40:24 +04:00
pmi->pmi_mountflags = (int)mountflags;
pmi->pmi_data = data;
pmi->pmi_uid = cred.sc_euid;
/*
* Connect to the remote socket if provided ans using SOCK_DGRAM
*/
if ((sock_type == SOCK_DGRAM) && sock) {
const struct sockaddr *sa;
struct sockaddr_un sun;
sa = (const struct sockaddr *)(void *)&sun;
sun.sun_len = sizeof(sun);
sun.sun_family = AF_LOCAL;
strcpy(sun.sun_path, sock);
if (connect(fd, sa, sun.sun_len) != 0)
DERR(EX_OSERR, "connect \"%s\" failed", sun.sun_path);
}
return;
}
static void
new_mount(int fd, int pmnt_flags)
{
struct puffs_usermount *pu;
struct perfuse_mount_info pmi;
struct perfuse_callbacks pc;
int ro_flag;
pid_t pid;
int flags;
int sock_type;
2012-01-29 10:22:01 +04:00
char trace_file[MAXPATHLEN + 1];
char trace_name[MAXPATHLEN + 1];
ssize_t trace_namelen;
int i;
pid = (perfuse_diagflags & PDF_FOREGROUND) ? 0 : fork();
switch(pid) {
case -1:
DERR(EX_OSERR, "cannot fork");
break;
case 0:
break;
default:
return;
/* NOTREACHED */
break;
}
/*
* Mount information (source, target, mount flags...)
*/
sock_type = pmnt_flags & PMNT_DGRAM ? SOCK_DGRAM : SOCK_SEQPACKET;
get_mount_info(fd, &pmi, sock_type);
/*
* Check that peer owns mountpoint and read (and write) on it?
*/
ro_flag = pmi.pmi_mountflags & MNT_RDONLY;
if (access_mount(pmi.pmi_target, pmi.pmi_uid, ro_flag) != 0)
DERRX(EX_NOPERM, "insuficient privileges to mount on %s",
pmi.pmi_target);
/*
* Initialize libperfuse, which will initialize libpuffs
*/
(void)memset(&pc, 0, sizeof(pc));
pc.pc_new_msg = perfused_new_pb;
pc.pc_xchg_msg = perfused_xchg_pb;
pc.pc_destroy_msg = (perfuse_destroy_msg_fn)puffs_framebuf_destroy;
pc.pc_get_inhdr = perfused_get_inhdr;
pc.pc_get_inpayload = perfused_get_inpayload;
pc.pc_get_outhdr = perfused_get_outhdr;
pc.pc_get_outpayload = perfused_get_outpayload;
pc.pc_umount = perfused_umount;
- Fix same vnodes associated with multiple cookies The scheme used to retreive known nodes on lookup was flawed, as it only used parent and name. This produced a different cookie for the same file if it was renamed, when looking up ../ or when dealing with multiple files associated with the same name through link(2). We therefore abandon the use of node name and introduce hashed lists of inodes. This causes a huge rewrite of reclaim code, which do not attempt to keep parents allocated until all their children are reclaimed - Fix race conditions in reclaim There are a few situations where we issue multiple FUSE operations for a PUFFS operation. On reclaim, we therefore have to wait for all FUSE operation to complete, not just the current exchanges. We do this by introducing node reference count with node_ref() and node_rele(). - Detect data loss caused by FAF VOP_PUTPAGES causes FAF writes where the kernel does not check the operation result. At least issue a warning on error. - Enjoy FAF shortcut on setattr No need to wait for the result if the kernel does not want it. There is however an exception for setattr that touch the size, we need to wait for completion because we have other operations queued for after the resize. - Fix fchmod() on write-open file fchmod() on a node open with write privilege will send setattr with both mode and size set. This confuses some FUSE filesystem. Therefore we send two FUSE operations, one for mode, and one for size. - Remove node TTL handling for netbsd-5 for simplicity sake. The code still builds on netbsd-5 but does not have the node TTL feature anymore. It works fine with kernel support on netbsd-6.
2012-07-21 09:49:42 +04:00
pc.pc_fsreq = perfused_gotframe;
pu = perfuse_init(&pc, &pmi);
puffs_framev_init(pu, perfused_readframe, perfused_writeframe,
- Fix same vnodes associated with multiple cookies The scheme used to retreive known nodes on lookup was flawed, as it only used parent and name. This produced a different cookie for the same file if it was renamed, when looking up ../ or when dealing with multiple files associated with the same name through link(2). We therefore abandon the use of node name and introduce hashed lists of inodes. This causes a huge rewrite of reclaim code, which do not attempt to keep parents allocated until all their children are reclaimed - Fix race conditions in reclaim There are a few situations where we issue multiple FUSE operations for a PUFFS operation. On reclaim, we therefore have to wait for all FUSE operation to complete, not just the current exchanges. We do this by introducing node reference count with node_ref() and node_rele(). - Detect data loss caused by FAF VOP_PUTPAGES causes FAF writes where the kernel does not check the operation result. At least issue a warning on error. - Enjoy FAF shortcut on setattr No need to wait for the result if the kernel does not want it. There is however an exception for setattr that touch the size, we need to wait for completion because we have other operations queued for after the resize. - Fix fchmod() on write-open file fchmod() on a node open with write privilege will send setattr with both mode and size set. This confuses some FUSE filesystem. Therefore we send two FUSE operations, one for mode, and one for size. - Remove node TTL handling for netbsd-5 for simplicity sake. The code still builds on netbsd-5 but does not have the node TTL feature anymore. It works fine with kernel support on netbsd-6.
2012-07-21 09:49:42 +04:00
perfused_cmpframe, pc.pc_fsreq, perfused_fdnotify);
if (puffs_framev_addfd(pu, fd, PUFFS_FBIO_READ|PUFFS_FBIO_WRITE) == -1)
DERR(EX_SOFTWARE, "puffs_framev_addfd failed");
perfuse_setspecific(pu, (void *)(intptr_t)fd);
setproctitle("perfused %s", pmi.pmi_target);
(void)kill(getpid(), SIGINFO); /* This is for -s option */
perfuse_fs_init(pu);
/*
* Non blocking I/O on /dev/fuse
* This must be done after perfused_fs_init
*/
if ((flags = fcntl(fd, F_GETFL, 0)) == -1)
DERR(EX_OSERR, "fcntl failed");
if (fcntl(fd, F_SETFL, flags|O_NONBLOCK) != 0)
DERR(EX_OSERR, "fcntl failed");
/*
* Setup trace file facility
*/
perfused_mount = pu;
2012-01-29 10:22:01 +04:00
trace_namelen = strlcpy(trace_name, pmi.pmi_target, MAXPATHLEN);
for (i = 0; i < trace_namelen; i++)
if (trace_name[i] == '/')
trace_name[i] = '-';
(void)snprintf(trace_file, MAXPATHLEN, _PATH_VAR_RUN_PERFUSE_TRACE,
trace_name);
if ((perfused_trace = fopen(trace_file, "w")) == NULL)
DERR(EX_OSFILE,
"could not open \"%s\"",
_PATH_VAR_RUN_PERFUSE_TRACE);
if (signal(SIGUSR1, sigusr1_handler) == SIG_ERR)
DERR(EX_OSERR, "signal failed");
/*
* Hand over control to puffs main loop.
*/
if (perfuse_mainloop(pu) != 0)
DERRX(EX_SOFTWARE, "perfused_mainloop exit");
/*
* Normal return after unmount
*/
return;
}
static int
parse_debug(char *optstr)
{
int retval = PDF_SYSLOG;
char *opt;
char *lastp;
for (opt = strtok_r(optstr, ",", &lastp);
opt;
opt = strtok_r(NULL, ",", &lastp)) {
if (strcmp(opt, "fuse") == 0)
retval |= PDF_FUSE;
else if (strcmp(opt, "puffs") == 0)
retval |= PDF_PUFFS;
else if (strcmp(opt, "dump") == 0)
retval |= PDF_DUMP;
else if (strcmp(opt, "fh") == 0)
retval |= PDF_FH;
else if (strcmp(opt, "readdir") == 0)
retval |= PDF_READDIR;
else if (strcmp(opt, "reclaim") == 0)
retval |= PDF_RECLAIM;
else if (strcmp(opt, "requeue") == 0)
retval |= PDF_REQUEUE;
else if (strcmp(opt, "sync") == 0)
retval |= PDF_SYNC;
else if (strcmp(opt, "misc") == 0)
retval |= PDF_MISC;
else if (strcmp(opt, "filename") == 0)
retval |= PDF_FILENAME;
else if (strcmp(opt, "reize") == 0)
retval |= PDF_RESIZE;
else
DWARNX("unknown debug flag \"%s\"", opt);
}
return retval;
}
/* ARGSUSED0 */
static void
siginfo_handler(int sig)
{
static int old_flags = 0;
int swap;
swap = perfuse_diagflags;
perfuse_diagflags = old_flags;
old_flags = swap;
DWARNX("debug %sabled", old_flags == 0 ? "en" : "dis");
return;
}
/* ARGSUSED0 */
static void
sigusr1_handler(int sig)
{
2012-01-29 10:22:01 +04:00
if (perfuse_diagflags & PDF_TRACE) {
perfuse_trace_dump(perfused_mount, perfused_trace);
2012-01-29 10:22:01 +04:00
perfuse_diagflags &= ~PDF_TRACE;
- Fix same vnodes associated with multiple cookies The scheme used to retreive known nodes on lookup was flawed, as it only used parent and name. This produced a different cookie for the same file if it was renamed, when looking up ../ or when dealing with multiple files associated with the same name through link(2). We therefore abandon the use of node name and introduce hashed lists of inodes. This causes a huge rewrite of reclaim code, which do not attempt to keep parents allocated until all their children are reclaimed - Fix race conditions in reclaim There are a few situations where we issue multiple FUSE operations for a PUFFS operation. On reclaim, we therefore have to wait for all FUSE operation to complete, not just the current exchanges. We do this by introducing node reference count with node_ref() and node_rele(). - Detect data loss caused by FAF VOP_PUTPAGES causes FAF writes where the kernel does not check the operation result. At least issue a warning on error. - Enjoy FAF shortcut on setattr No need to wait for the result if the kernel does not want it. There is however an exception for setattr that touch the size, we need to wait for completion because we have other operations queued for after the resize. - Fix fchmod() on write-open file fchmod() on a node open with write privilege will send setattr with both mode and size set. This confuses some FUSE filesystem. Therefore we send two FUSE operations, one for mode, and one for size. - Remove node TTL handling for netbsd-5 for simplicity sake. The code still builds on netbsd-5 but does not have the node TTL feature anymore. It works fine with kernel support on netbsd-6.
2012-07-21 09:49:42 +04:00
DPRINTF("trace dumped, trace disabled\n");
2012-01-29 10:22:01 +04:00
} else {
perfuse_diagflags |= PDF_TRACE;
- Fix same vnodes associated with multiple cookies The scheme used to retreive known nodes on lookup was flawed, as it only used parent and name. This produced a different cookie for the same file if it was renamed, when looking up ../ or when dealing with multiple files associated with the same name through link(2). We therefore abandon the use of node name and introduce hashed lists of inodes. This causes a huge rewrite of reclaim code, which do not attempt to keep parents allocated until all their children are reclaimed - Fix race conditions in reclaim There are a few situations where we issue multiple FUSE operations for a PUFFS operation. On reclaim, we therefore have to wait for all FUSE operation to complete, not just the current exchanges. We do this by introducing node reference count with node_ref() and node_rele(). - Detect data loss caused by FAF VOP_PUTPAGES causes FAF writes where the kernel does not check the operation result. At least issue a warning on error. - Enjoy FAF shortcut on setattr No need to wait for the result if the kernel does not want it. There is however an exception for setattr that touch the size, we need to wait for completion because we have other operations queued for after the resize. - Fix fchmod() on write-open file fchmod() on a node open with write privilege will send setattr with both mode and size set. This confuses some FUSE filesystem. Therefore we send two FUSE operations, one for mode, and one for size. - Remove node TTL handling for netbsd-5 for simplicity sake. The code still builds on netbsd-5 but does not have the node TTL feature anymore. It works fine with kernel support on netbsd-6.
2012-07-21 09:49:42 +04:00
DPRINTF("trace enabled\n");
2012-01-29 10:22:01 +04:00
}
return;
}
static int
parse_options(int argc, char **argv)
{
int ch;
int foreground = 0;
int retval = -1;
perfuse_diagflags = PDF_FOREGROUND | PDF_SYSLOG;
while ((ch = getopt(argc, argv, "d:fsi:")) != -1) {
switch (ch) {
case 'd':
perfuse_diagflags |= parse_debug(optarg);
break;
case 's':
if (signal(SIGINFO, siginfo_handler) == SIG_ERR)
DERR(EX_OSERR, "signal failed");
break;
case 'f':
foreground = 1;
perfuse_diagflags |= PDF_MISC;
break;
case 'i':
retval = atoi(optarg);
foreground = 1;
break;
default:
DERRX(EX_USAGE, "%s [-fs] [-d classes] [-i fd]", argv[0]);
break;
}
}
if (!foreground)
perfuse_diagflags &= ~PDF_FOREGROUND;
return retval;
}
int
main(int argc, char **argv)
{
int s;
int sock_type;
socklen_t len;
s = parse_options(argc, argv);
if (perfuse_diagflags & PDF_SYSLOG)
openlog("perfused", LOG_NDELAY, LOG_DAEMON);
if (!(perfuse_diagflags & PDF_FOREGROUND))
if (daemon(0, 0) != 0)
DERR(EX_OSERR, "daemon failed");
if (s != -1) {
new_mount(s, PMNT_SOCKPAIR);
exit(0);
}
s = perfused_open_sock();
#ifdef PERFUSE_DEBUG
if (perfuse_diagflags & PDF_MISC)
DPRINTF("perfused ready\n");
#endif
len = sizeof(sock_type);
if (getsockopt(s, SOL_SOCKET, SO_TYPE, &sock_type, &len) != 0)
DERR(EX_OSERR, "getsockopt SO_TYPE failed");
switch(sock_type) {
case SOCK_DGRAM:
new_mount(s, PMNT_DEVFUSE|PMNT_DGRAM);
exit(0);
case SOCK_SEQPACKET:
if (listen(s, 0) != 0)
DERR(EX_OSERR, "listen failed");
for (;;) {
int fd;
struct sockaddr_un sun;
struct sockaddr *sa;
len = sizeof(sun);
sa = (struct sockaddr *)(void *)&sun;
if ((fd = accept(s, sa, &len)) == -1)
DERR(EX_OSERR, "accept failed");
new_mount(fd, PMNT_DEVFUSE);
}
default:
DERRX(EX_SOFTWARE, "unexpected so_type %d", sock_type);
break;
}
/* NOTREACHED */
return 0;
}
void
perfused_panic(void)
{
DWARNX("filesystem crashed");
exit(EX_OK);
}