freebsd-dev/sys/nfs/bootp_subr.c

1826 lines
44 KiB
C
Raw Normal View History

/*-
* SPDX-License-Identifier: BSD-4-Clause
*
* Copyright (c) 1995 Gordon Ross, Adam Glass
* Copyright (c) 1992 Regents of the University of California.
* All rights reserved.
*
* This software was developed by the Computer Systems Engineering group
* at Lawrence Berkeley Laboratory under DARPA contract BG 91-66 and
* contributed to Berkeley.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* This product includes software developed by the University of
* California, Lawrence Berkeley Laboratory and its contributors.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* based on:
* nfs/krpc_subr.c
* $NetBSD: krpc_subr.c,v 1.10 1995/08/08 20:43:43 gwr Exp $
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_bootp.h"
#include "opt_nfs.h"
Changes to allow using BOOTP_NFSROOT and mounting an nfs root filesystem other than the one specified by the BOOTP server. This configures NFS using the BOOTP protocol while also respecting other root-path options such as setting vfs.root.mountfrom in the environment or using the RB_DFLTROOT boot option. It allows you to override the root path provided by the server, or to supply a root path when the server provides IP configuration but no root path info. This maintains the historical BOOTP_NFSROOT behavior of panicking on a failure to mount the root path provided by the server, unless you've provided an alternative via the ROOTDEVNAME kernel option or by setting vfs.root.mountfrom. The behavior of panicking when given no other options is preserved because it amounts to a bit of a retry loop that could eventually recover from a transient network or server problem. The user can now override the root path from loader(8) even if the kernel is compiled with BOOTP_NFSROOT. If vfs.root.mountfrom is set in the environment it is used unconditionally -- it always overrides the BOOTP info. If it begins with [old]nfs: then the BOOTP code uses it instead of the server-provided info. If it specifies some other filesystem then the bootp code will not panic like it used to and the code in vfs_mountroot.c will invoke the right filesystem to do the mount. If the kernel is compiled with the ROOTDEVNAME option, then that name is used by the BOOTP code if either * The server doesn't provide a pathname. * The boothowto flags include RB_DFLTROOT. The latter allows the user to compile in alternate path in ROOTDEVNAME such as ufs:/dev/da0s1a and boot from that path by setting boot_dftlroot=1 in loader(8) or using the '-r' option in boot(8). The one thing not provided here is automatic failover from a server-provided path to a compiled-in one without the user manually requesting that. The code just isn't currently structured in a way that makes that possible with a lot of rewrite. I think the ability to set vfs.root.mountfrom and to use ROOTDEVNAME automatically when the server doesn't provide a name covers the most common needs. A set of patches submitted by Lars Eggert provided the part I couldn't figure out by myself when I tried to do this last year; many thanks. Reviewed by: rodrigc
2013-07-31 19:14:00 +00:00
#include "opt_rootdevname.h"
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/endian.h>
#include <sys/jail.h>
#include <sys/kernel.h>
#include <sys/sockio.h>
#include <sys/malloc.h>
#include <sys/mount.h>
#include <sys/mbuf.h>
#include <sys/proc.h>
Changes to allow using BOOTP_NFSROOT and mounting an nfs root filesystem other than the one specified by the BOOTP server. This configures NFS using the BOOTP protocol while also respecting other root-path options such as setting vfs.root.mountfrom in the environment or using the RB_DFLTROOT boot option. It allows you to override the root path provided by the server, or to supply a root path when the server provides IP configuration but no root path info. This maintains the historical BOOTP_NFSROOT behavior of panicking on a failure to mount the root path provided by the server, unless you've provided an alternative via the ROOTDEVNAME kernel option or by setting vfs.root.mountfrom. The behavior of panicking when given no other options is preserved because it amounts to a bit of a retry loop that could eventually recover from a transient network or server problem. The user can now override the root path from loader(8) even if the kernel is compiled with BOOTP_NFSROOT. If vfs.root.mountfrom is set in the environment it is used unconditionally -- it always overrides the BOOTP info. If it begins with [old]nfs: then the BOOTP code uses it instead of the server-provided info. If it specifies some other filesystem then the bootp code will not panic like it used to and the code in vfs_mountroot.c will invoke the right filesystem to do the mount. If the kernel is compiled with the ROOTDEVNAME option, then that name is used by the BOOTP code if either * The server doesn't provide a pathname. * The boothowto flags include RB_DFLTROOT. The latter allows the user to compile in alternate path in ROOTDEVNAME such as ufs:/dev/da0s1a and boot from that path by setting boot_dftlroot=1 in loader(8) or using the '-r' option in boot(8). The one thing not provided here is automatic failover from a server-provided path to a compiled-in one without the user manually requesting that. The code just isn't currently structured in a way that makes that possible with a lot of rewrite. I think the ability to set vfs.root.mountfrom and to use ROOTDEVNAME automatically when the server doesn't provide a name covers the most common needs. A set of patches submitted by Lars Eggert provided the part I couldn't figure out by myself when I tried to do this last year; many thanks. Reviewed by: rodrigc
2013-07-31 19:14:00 +00:00
#include <sys/reboot.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/sysctl.h>
#include <sys/uio.h>
#include <net/if.h>
#include <net/if_var.h>
#include <net/route.h>
#include <net/route/route_ctl.h>
#include <netinet/in.h>
#include <netinet/in_var.h>
#include <net/if_types.h>
#include <net/if_dl.h>
#include <net/vnet.h>
#include <nfs/nfsproto.h>
#include <nfsclient/nfs.h>
#include <nfs/nfsdiskless.h>
#include <nfs/krpc.h>
#include <nfs/xdr_subs.h>
#define BOOTP_MIN_LEN 300 /* Minimum size of bootp udp packet */
#ifndef BOOTP_SETTLE_DELAY
#define BOOTP_SETTLE_DELAY 3
#endif
/*
* Wait 10 seconds for interface appearance
* USB ethernet adapters might require some time to pop up
*/
#ifndef BOOTP_IFACE_WAIT_TIMEOUT
#define BOOTP_IFACE_WAIT_TIMEOUT 10
#endif
/*
* What is the longest we will wait before re-sending a request?
* Note this is also the frequency of "RPC timeout" messages.
* The re-send loop count sup linearly to this maximum, so the
* first complaint will happen after (1+2+3+4+5)=15 seconds.
*/
#define MAX_RESEND_DELAY 5 /* seconds */
/* Definitions from RFC951 */
struct bootp_packet {
2000-10-24 22:40:18 +00:00
u_int8_t op;
u_int8_t htype;
u_int8_t hlen;
u_int8_t hops;
u_int32_t xid;
u_int16_t secs;
u_int16_t flags;
struct in_addr ciaddr;
struct in_addr yiaddr;
struct in_addr siaddr;
struct in_addr giaddr;
unsigned char chaddr[16];
char sname[64];
char file[128];
unsigned char vend[1222];
};
struct bootpc_ifcontext {
STAILQ_ENTRY(bootpc_ifcontext) next;
struct bootp_packet call;
struct bootp_packet reply;
int replylen;
int overload;
union {
struct ifreq _ifreq;
struct in_aliasreq _in_alias_req;
} _req;
#define ireq _req._ifreq
#define iareq _req._in_alias_req
struct ifnet *ifp;
struct sockaddr_dl *sdl;
struct sockaddr_in myaddr;
struct sockaddr_in netmask;
struct sockaddr_in gw;
int gotgw;
int gotnetmask;
int gotrootpath;
int outstanding;
int sentmsg;
u_int32_t xid;
enum {
IF_BOOTP_UNRESOLVED,
IF_BOOTP_RESOLVED,
IF_BOOTP_FAILED,
IF_DHCP_UNRESOLVED,
IF_DHCP_OFFERED,
IF_DHCP_RESOLVED,
IF_DHCP_FAILED,
} state;
int dhcpquerytype; /* dhcp type sent */
struct in_addr dhcpserver;
int gotdhcpserver;
uint16_t mtu;
};
#define TAG_MAXLEN 1024
struct bootpc_tagcontext {
char buf[TAG_MAXLEN + 1];
int overload;
int badopt;
int badtag;
int foundopt;
int taglen;
};
struct bootpc_globalcontext {
STAILQ_HEAD(, bootpc_ifcontext) interfaces;
u_int32_t xid;
Changes to allow using BOOTP_NFSROOT and mounting an nfs root filesystem other than the one specified by the BOOTP server. This configures NFS using the BOOTP protocol while also respecting other root-path options such as setting vfs.root.mountfrom in the environment or using the RB_DFLTROOT boot option. It allows you to override the root path provided by the server, or to supply a root path when the server provides IP configuration but no root path info. This maintains the historical BOOTP_NFSROOT behavior of panicking on a failure to mount the root path provided by the server, unless you've provided an alternative via the ROOTDEVNAME kernel option or by setting vfs.root.mountfrom. The behavior of panicking when given no other options is preserved because it amounts to a bit of a retry loop that could eventually recover from a transient network or server problem. The user can now override the root path from loader(8) even if the kernel is compiled with BOOTP_NFSROOT. If vfs.root.mountfrom is set in the environment it is used unconditionally -- it always overrides the BOOTP info. If it begins with [old]nfs: then the BOOTP code uses it instead of the server-provided info. If it specifies some other filesystem then the bootp code will not panic like it used to and the code in vfs_mountroot.c will invoke the right filesystem to do the mount. If the kernel is compiled with the ROOTDEVNAME option, then that name is used by the BOOTP code if either * The server doesn't provide a pathname. * The boothowto flags include RB_DFLTROOT. The latter allows the user to compile in alternate path in ROOTDEVNAME such as ufs:/dev/da0s1a and boot from that path by setting boot_dftlroot=1 in loader(8) or using the '-r' option in boot(8). The one thing not provided here is automatic failover from a server-provided path to a compiled-in one without the user manually requesting that. The code just isn't currently structured in a way that makes that possible with a lot of rewrite. I think the ability to set vfs.root.mountfrom and to use ROOTDEVNAME automatically when the server doesn't provide a name covers the most common needs. A set of patches submitted by Lars Eggert provided the part I couldn't figure out by myself when I tried to do this last year; many thanks. Reviewed by: rodrigc
2013-07-31 19:14:00 +00:00
int any_root_overrides;
int gotrootpath;
int gotgw;
int ifnum;
int secs;
int starttime;
struct bootp_packet reply;
int replylen;
struct bootpc_ifcontext *setrootfs;
struct bootpc_ifcontext *sethostname;
struct bootpc_tagcontext tmptag;
struct bootpc_tagcontext tag;
};
#define IPPORT_BOOTPC 68
#define IPPORT_BOOTPS 67
#define BOOTP_REQUEST 1
#define BOOTP_REPLY 2
/* Common tags */
#define TAG_PAD 0 /* Pad option, implicit length 1 */
#define TAG_SUBNETMASK 1 /* RFC 950 subnet mask */
#define TAG_ROUTERS 3 /* Routers (in order of preference) */
#define TAG_HOSTNAME 12 /* Client host name */
#define TAG_ROOT 17 /* Root path */
#define TAG_INTF_MTU 26 /* Interface MTU Size (RFC2132) */
/* DHCP specific tags */
#define TAG_OVERLOAD 52 /* Option Overload */
#define TAG_MAXMSGSIZE 57 /* Maximum DHCP Message Size */
#define TAG_END 255 /* End Option (i.e. no more options) */
/* Overload values */
#define OVERLOAD_FILE 1
#define OVERLOAD_SNAME 2
/* Site specific tags: */
#define TAG_ROOTOPTS 130
#define TAG_COOKIE 134 /* ascii info for userland, via sysctl */
#define TAG_DHCP_MSGTYPE 53
#define TAG_DHCP_REQ_ADDR 50
#define TAG_DHCP_SERVERID 54
#define TAG_DHCP_LEASETIME 51
#define TAG_VENDOR_INDENTIFIER 60
#define DHCP_NOMSG 0
#define DHCP_DISCOVER 1
#define DHCP_OFFER 2
#define DHCP_REQUEST 3
#define DHCP_ACK 5
/* NFS read/write block size */
#ifndef BOOTP_BLOCKSIZE
#define BOOTP_BLOCKSIZE 8192
#endif
static char bootp_cookie[128];
static struct socket *bootp_so;
SYSCTL_STRING(_kern, OID_AUTO, bootp_cookie, CTLFLAG_RD,
bootp_cookie, 0, "Cookie (T134) supplied by bootp server");
/* mountd RPC */
static int md_mount(struct sockaddr_in *mdsin, char *path, u_char *fhp,
int *fhsizep, struct nfs_args *args, struct thread *td);
static int setfs(struct sockaddr_in *addr, char *path, char *p,
const struct in_addr *siaddr);
static int getdec(char **ptr);
static int getip(char **ptr, struct in_addr *ip);
static void mountopts(struct nfs_args *args, char *p);
static int xdr_opaque_decode(struct mbuf **ptr, u_char *buf, int len);
static int xdr_int_decode(struct mbuf **ptr, int *iptr);
static void print_in_addr(struct in_addr addr);
static void print_sin_addr(struct sockaddr_in *addr);
static void clear_sinaddr(struct sockaddr_in *sin);
static void allocifctx(struct bootpc_globalcontext *gctx);
static void bootpc_compose_query(struct bootpc_ifcontext *ifctx,
struct thread *td);
static unsigned char *bootpc_tag(struct bootpc_tagcontext *tctx,
struct bootp_packet *bp, int len, int tag);
static void bootpc_tag_helper(struct bootpc_tagcontext *tctx,
unsigned char *start, int len, int tag);
#ifdef BOOTP_DEBUG
void bootpboot_p_if(struct ifnet *ifp, struct ifaddr *ifa);
void bootpboot_p_iflist(void);
#endif
static int bootpc_call(struct bootpc_globalcontext *gctx,
struct thread *td);
static void bootpc_fakeup_interface(struct bootpc_ifcontext *ifctx,
struct thread *td);
static void bootpc_adjust_interface(struct bootpc_ifcontext *ifctx,
struct bootpc_globalcontext *gctx, struct thread *td);
static void bootpc_decode_reply(struct nfsv3_diskless *nd,
struct bootpc_ifcontext *ifctx,
struct bootpc_globalcontext *gctx);
static int bootpc_received(struct bootpc_globalcontext *gctx,
struct bootpc_ifcontext *ifctx);
static __inline int bootpc_ifctx_isresolved(struct bootpc_ifcontext *ifctx);
static __inline int bootpc_ifctx_isunresolved(struct bootpc_ifcontext *ifctx);
static __inline int bootpc_ifctx_isfailed(struct bootpc_ifcontext *ifctx);
/*
* In order to have multiple active interfaces with address 0.0.0.0
* and be able to send data to a selected interface, we first set
* mask to /8 on all interfaces, and temporarily set it to /0 when
* doing sosend().
*/
#ifdef BOOTP_DEBUG
void
bootpboot_p_if(struct ifnet *ifp, struct ifaddr *ifa)
{
printf("%s flags %x, addr ",
ifp->if_xname, ifp->if_flags);
print_sin_addr((struct sockaddr_in *) ifa->ifa_addr);
printf(", broadcast ");
print_sin_addr((struct sockaddr_in *) ifa->ifa_dstaddr);
printf(", netmask ");
print_sin_addr((struct sockaddr_in *) ifa->ifa_netmask);
printf("\n");
}
2000-10-24 22:40:18 +00:00
void
bootpboot_p_iflist(void)
{
2000-10-24 22:40:18 +00:00
struct ifnet *ifp;
struct ifaddr *ifa;
2000-10-24 22:40:18 +00:00
printf("Interface list:\n");
IFNET_RLOCK();
for (ifp = CK_STAILQ_FIRST(&V_ifnet);
2000-10-24 22:40:18 +00:00
ifp != NULL;
ifp = CK_STAILQ_NEXT(ifp, if_link)) {
ifnet: Replace if_addr_lock rwlock with epoch + mutex Run on LLNW canaries and tested by pho@ gallatin: Using a 14-core, 28-HTT single socket E5-2697 v3 with a 40GbE MLX5 based ConnectX 4-LX NIC, I see an almost 12% improvement in received packet rate, and a larger improvement in bytes delivered all the way to userspace. When the host receiving 64 streams of netperf -H $DUT -t UDP_STREAM -- -m 1, I see, using nstat -I mce0 1 before the patch: InMpps OMpps InGbs OGbs err TCP Est %CPU syscalls csw irq GBfree 4.98 0.00 4.42 0.00 4235592 33 83.80 4720653 2149771 1235 247.32 4.73 0.00 4.20 0.00 4025260 33 82.99 4724900 2139833 1204 247.32 4.72 0.00 4.20 0.00 4035252 33 82.14 4719162 2132023 1264 247.32 4.71 0.00 4.21 0.00 4073206 33 83.68 4744973 2123317 1347 247.32 4.72 0.00 4.21 0.00 4061118 33 80.82 4713615 2188091 1490 247.32 4.72 0.00 4.21 0.00 4051675 33 85.29 4727399 2109011 1205 247.32 4.73 0.00 4.21 0.00 4039056 33 84.65 4724735 2102603 1053 247.32 After the patch InMpps OMpps InGbs OGbs err TCP Est %CPU syscalls csw irq GBfree 5.43 0.00 4.20 0.00 3313143 33 84.96 5434214 1900162 2656 245.51 5.43 0.00 4.20 0.00 3308527 33 85.24 5439695 1809382 2521 245.51 5.42 0.00 4.19 0.00 3316778 33 87.54 5416028 1805835 2256 245.51 5.42 0.00 4.19 0.00 3317673 33 90.44 5426044 1763056 2332 245.51 5.42 0.00 4.19 0.00 3314839 33 88.11 5435732 1792218 2499 245.52 5.44 0.00 4.19 0.00 3293228 33 91.84 5426301 1668597 2121 245.52 Similarly, netperf reports 230Mb/s before the patch, and 270Mb/s after the patch Reviewed by: gallatin Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15366
2018-05-18 20:13:34 +00:00
for (ifa = CK_STAILQ_FIRST(&ifp->if_addrhead);
2000-10-24 22:40:18 +00:00
ifa != NULL;
ifa = CK_STAILQ_NEXT(ifa, ifa_link))
if (ifa->ifa_addr->sa_family == AF_INET)
bootpboot_p_if(ifp, ifa);
}
2002-12-22 05:35:03 +00:00
IFNET_RUNLOCK();
}
#endif /* defined(BOOTP_DEBUG) */
static void
clear_sinaddr(struct sockaddr_in *sin)
{
bzero(sin, sizeof(*sin));
sin->sin_len = sizeof(*sin);
sin->sin_family = AF_INET;
sin->sin_addr.s_addr = INADDR_ANY; /* XXX: htonl(INAADDR_ANY) ? */
sin->sin_port = 0;
}
static void
allocifctx(struct bootpc_globalcontext *gctx)
{
struct bootpc_ifcontext *ifctx;
ifctx = malloc(sizeof(*ifctx), M_TEMP, M_WAITOK | M_ZERO);
ifctx->xid = gctx->xid;
#ifdef BOOTP_NO_DHCP
ifctx->state = IF_BOOTP_UNRESOLVED;
#else
ifctx->state = IF_DHCP_UNRESOLVED;
#endif
gctx->xid += 0x100;
STAILQ_INSERT_TAIL(&gctx->interfaces, ifctx, next);
}
static __inline int
bootpc_ifctx_isresolved(struct bootpc_ifcontext *ifctx)
{
if (ifctx->state == IF_BOOTP_RESOLVED ||
ifctx->state == IF_DHCP_RESOLVED)
return 1;
return 0;
}
static __inline int
bootpc_ifctx_isunresolved(struct bootpc_ifcontext *ifctx)
{
if (ifctx->state == IF_BOOTP_UNRESOLVED ||
ifctx->state == IF_DHCP_UNRESOLVED)
return 1;
return 0;
}
static __inline int
bootpc_ifctx_isfailed(struct bootpc_ifcontext *ifctx)
{
if (ifctx->state == IF_BOOTP_FAILED ||
ifctx->state == IF_DHCP_FAILED)
return 1;
return 0;
}
static int
bootpc_received(struct bootpc_globalcontext *gctx,
struct bootpc_ifcontext *ifctx)
{
unsigned char dhcpreplytype;
char *p;
/*
* Need timeout for fallback to less
* desirable alternative.
*/
/* This call used for the side effect (badopt flag) */
(void) bootpc_tag(&gctx->tmptag, &gctx->reply,
gctx->replylen,
TAG_END);
/* If packet is invalid, ignore it */
if (gctx->tmptag.badopt != 0)
return 0;
p = bootpc_tag(&gctx->tmptag, &gctx->reply,
gctx->replylen, TAG_DHCP_MSGTYPE);
if (p != NULL)
dhcpreplytype = *p;
else
dhcpreplytype = DHCP_NOMSG;
switch (ifctx->dhcpquerytype) {
case DHCP_DISCOVER:
if (dhcpreplytype != DHCP_OFFER /* Normal DHCP offer */
#ifndef BOOTP_FORCE_DHCP
&& dhcpreplytype != DHCP_NOMSG /* Fallback to BOOTP */
#endif
)
return 0;
break;
case DHCP_REQUEST:
if (dhcpreplytype != DHCP_ACK)
return 0;
case DHCP_NOMSG:
break;
}
/* Ignore packet unless it gives us a root tag we didn't have */
if ((ifctx->state == IF_BOOTP_RESOLVED ||
(ifctx->dhcpquerytype == DHCP_DISCOVER &&
(ifctx->state == IF_DHCP_OFFERED ||
ifctx->state == IF_DHCP_RESOLVED))) &&
(bootpc_tag(&gctx->tmptag, &ifctx->reply,
ifctx->replylen,
TAG_ROOT) != NULL ||
bootpc_tag(&gctx->tmptag, &gctx->reply,
gctx->replylen,
TAG_ROOT) == NULL))
return 0;
bcopy(&gctx->reply, &ifctx->reply, gctx->replylen);
ifctx->replylen = gctx->replylen;
/* XXX: Only reset if 'perfect' response */
if (ifctx->state == IF_BOOTP_UNRESOLVED)
ifctx->state = IF_BOOTP_RESOLVED;
else if (ifctx->state == IF_DHCP_UNRESOLVED &&
ifctx->dhcpquerytype == DHCP_DISCOVER) {
if (dhcpreplytype == DHCP_OFFER)
ifctx->state = IF_DHCP_OFFERED;
else
ifctx->state = IF_BOOTP_RESOLVED; /* Fallback */
} else if (ifctx->state == IF_DHCP_OFFERED &&
ifctx->dhcpquerytype == DHCP_REQUEST)
ifctx->state = IF_DHCP_RESOLVED;
if (ifctx->dhcpquerytype == DHCP_DISCOVER &&
ifctx->state != IF_BOOTP_RESOLVED) {
p = bootpc_tag(&gctx->tmptag, &ifctx->reply,
ifctx->replylen, TAG_DHCP_SERVERID);
if (p != NULL && gctx->tmptag.taglen == 4) {
memcpy(&ifctx->dhcpserver, p, 4);
ifctx->gotdhcpserver = 1;
} else
ifctx->gotdhcpserver = 0;
return 1;
}
ifctx->gotrootpath = (bootpc_tag(&gctx->tmptag, &ifctx->reply,
ifctx->replylen,
TAG_ROOT) != NULL);
ifctx->gotgw = (bootpc_tag(&gctx->tmptag, &ifctx->reply,
ifctx->replylen,
TAG_ROUTERS) != NULL);
ifctx->gotnetmask = (bootpc_tag(&gctx->tmptag, &ifctx->reply,
ifctx->replylen,
TAG_SUBNETMASK) != NULL);
return 1;
}
1998-02-09 06:11:36 +00:00
static int
bootpc_call(struct bootpc_globalcontext *gctx, struct thread *td)
{
struct sockaddr_in *sin, dst;
struct uio auio;
struct sockopt sopt;
struct iovec aio;
int error, on, rcvflg, timo, len;
time_t atimo;
time_t rtimo;
struct timeval tv;
struct bootpc_ifcontext *ifctx;
int outstanding;
int gotrootpath;
int retry;
const char *s;
tv.tv_sec = 1;
tv.tv_usec = 0;
bzero(&sopt, sizeof(sopt));
sopt.sopt_dir = SOPT_SET;
sopt.sopt_level = SOL_SOCKET;
sopt.sopt_name = SO_RCVTIMEO;
sopt.sopt_val = &tv;
sopt.sopt_valsize = sizeof tv;
error = sosetopt(bootp_so, &sopt);
if (error != 0)
goto out;
/*
* Enable broadcast.
*/
on = 1;
sopt.sopt_name = SO_BROADCAST;
sopt.sopt_val = &on;
sopt.sopt_valsize = sizeof on;
error = sosetopt(bootp_so, &sopt);
if (error != 0)
goto out;
/*
* Disable routing.
*/
on = 1;
sopt.sopt_name = SO_DONTROUTE;
sopt.sopt_val = &on;
sopt.sopt_valsize = sizeof on;
error = sosetopt(bootp_so, &sopt);
if (error != 0)
goto out;
/*
* Bind the local endpoint to a bootp client port.
*/
sin = &dst;
clear_sinaddr(sin);
sin->sin_port = htons(IPPORT_BOOTPC);
error = sobind(bootp_so, (struct sockaddr *)sin, td);
if (error != 0) {
printf("bind failed\n");
goto out;
}
/*
* Setup socket address for the server.
*/
sin = &dst;
clear_sinaddr(sin);
sin->sin_addr.s_addr = INADDR_BROADCAST;
sin->sin_port = htons(IPPORT_BOOTPS);
/*
* Send it, repeatedly, until a reply is received,
* but delay each re-send by an increasing amount.
* If the delay hits the maximum, start complaining.
*/
timo = 0;
rtimo = 0;
for (;;) {
outstanding = 0;
gotrootpath = 0;
STAILQ_FOREACH(ifctx, &gctx->interfaces, next) {
if (bootpc_ifctx_isresolved(ifctx) != 0 &&
bootpc_tag(&gctx->tmptag, &ifctx->reply,
ifctx->replylen,
TAG_ROOT) != NULL)
gotrootpath = 1;
}
STAILQ_FOREACH(ifctx, &gctx->interfaces, next) {
struct in_aliasreq *ifra = &ifctx->iareq;
sin = (struct sockaddr_in *)&ifra->ifra_mask;
ifctx->outstanding = 0;
if (bootpc_ifctx_isresolved(ifctx) != 0 &&
gotrootpath != 0) {
continue;
}
if (bootpc_ifctx_isfailed(ifctx) != 0)
continue;
outstanding++;
ifctx->outstanding = 1;
/* Proceed to next step in DHCP negotiation */
if ((ifctx->state == IF_DHCP_OFFERED &&
ifctx->dhcpquerytype != DHCP_REQUEST) ||
(ifctx->state == IF_DHCP_UNRESOLVED &&
ifctx->dhcpquerytype != DHCP_DISCOVER) ||
(ifctx->state == IF_BOOTP_UNRESOLVED &&
ifctx->dhcpquerytype != DHCP_NOMSG)) {
ifctx->sentmsg = 0;
bootpc_compose_query(ifctx, td);
}
/* Send BOOTP request (or re-send). */
if (ifctx->sentmsg == 0) {
switch(ifctx->dhcpquerytype) {
case DHCP_DISCOVER:
s = "DHCP Discover";
break;
case DHCP_REQUEST:
s = "DHCP Request";
break;
case DHCP_NOMSG:
default:
s = "BOOTP Query";
break;
}
printf("Sending %s packet from "
"interface %s (%*D)\n",
s,
ifctx->ireq.ifr_name,
ifctx->sdl->sdl_alen,
(unsigned char *) LLADDR(ifctx->sdl),
":");
ifctx->sentmsg = 1;
}
aio.iov_base = (caddr_t) &ifctx->call;
aio.iov_len = sizeof(ifctx->call);
auio.uio_iov = &aio;
auio.uio_iovcnt = 1;
auio.uio_segflg = UIO_SYSSPACE;
auio.uio_rw = UIO_WRITE;
auio.uio_offset = 0;
auio.uio_resid = sizeof(ifctx->call);
auio.uio_td = td;
/* Set netmask to 0.0.0.0 */
clear_sinaddr(sin);
error = ifioctl(bootp_so, SIOCAIFADDR, (caddr_t)ifra,
td);
if (error != 0)
panic("%s: SIOCAIFADDR, error=%d", __func__,
error);
error = sosend(bootp_so, (struct sockaddr *) &dst,
&auio, NULL, NULL, 0, td);
if (error != 0)
printf("%s: sosend: %d state %08x\n", __func__,
error, (int )bootp_so->so_state);
/* Set netmask to 255.0.0.0 */
sin->sin_addr.s_addr = htonl(IN_CLASSA_NET);
error = ifioctl(bootp_so, SIOCAIFADDR, (caddr_t)ifra,
td);
if (error != 0)
panic("%s: SIOCAIFADDR, error=%d", __func__,
error);
}
if (outstanding == 0 &&
(rtimo == 0 || time_second >= rtimo)) {
error = 0;
goto out;
}
/* Determine new timeout. */
if (timo < MAX_RESEND_DELAY)
timo++;
else {
printf("DHCP/BOOTP timeout for server ");
print_sin_addr(&dst);
printf("\n");
}
/*
* Wait for up to timo seconds for a reply.
* The socket receive timeout was set to 1 second.
*/
atimo = timo + time_second;
while (time_second < atimo) {
aio.iov_base = (caddr_t) &gctx->reply;
aio.iov_len = sizeof(gctx->reply);
auio.uio_iov = &aio;
auio.uio_iovcnt = 1;
auio.uio_segflg = UIO_SYSSPACE;
auio.uio_rw = UIO_READ;
auio.uio_offset = 0;
auio.uio_resid = sizeof(gctx->reply);
auio.uio_td = td;
rcvflg = 0;
error = soreceive(bootp_so, NULL, &auio,
2000-10-24 22:40:18 +00:00
NULL, NULL, &rcvflg);
gctx->secs = time_second - gctx->starttime;
STAILQ_FOREACH(ifctx, &gctx->interfaces, next) {
if (bootpc_ifctx_isresolved(ifctx) != 0 ||
bootpc_ifctx_isfailed(ifctx) != 0)
continue;
ifctx->call.secs = htons(gctx->secs);
}
if (error == EWOULDBLOCK)
continue;
if (error != 0)
goto out;
len = sizeof(gctx->reply) - auio.uio_resid;
/* Do we have the required number of bytes ? */
if (len < BOOTP_MIN_LEN)
continue;
gctx->replylen = len;
/* Is it a reply? */
if (gctx->reply.op != BOOTP_REPLY)
2000-10-24 22:40:18 +00:00
continue;
/* Is this an answer to our query */
STAILQ_FOREACH(ifctx, &gctx->interfaces, next) {
if (gctx->reply.xid != ifctx->call.xid)
continue;
/* Same HW address size ? */
if (gctx->reply.hlen != ifctx->call.hlen)
continue;
/* Correct HW address ? */
if (bcmp(gctx->reply.chaddr,
ifctx->call.chaddr,
ifctx->call.hlen) != 0)
continue;
break;
}
if (ifctx != NULL) {
s = bootpc_tag(&gctx->tmptag,
&gctx->reply,
gctx->replylen,
TAG_DHCP_MSGTYPE);
if (s != NULL) {
switch (*s) {
case DHCP_OFFER:
s = "DHCP Offer";
break;
case DHCP_ACK:
s = "DHCP Ack";
break;
default:
s = "DHCP (unexpected)";
break;
}
} else
s = "BOOTP Reply";
printf("Received %s packet"
" on %s from ",
s,
ifctx->ireq.ifr_name);
print_in_addr(gctx->reply.siaddr);
if (gctx->reply.giaddr.s_addr !=
htonl(INADDR_ANY)) {
printf(" via ");
print_in_addr(gctx->reply.giaddr);
}
if (bootpc_received(gctx, ifctx) != 0) {
printf(" (accepted)");
if (ifctx->outstanding) {
ifctx->outstanding = 0;
outstanding--;
}
/* Network settle delay */
if (outstanding == 0)
atimo = time_second +
BOOTP_SETTLE_DELAY;
} else
printf(" (ignored)");
Changes to allow using BOOTP_NFSROOT and mounting an nfs root filesystem other than the one specified by the BOOTP server. This configures NFS using the BOOTP protocol while also respecting other root-path options such as setting vfs.root.mountfrom in the environment or using the RB_DFLTROOT boot option. It allows you to override the root path provided by the server, or to supply a root path when the server provides IP configuration but no root path info. This maintains the historical BOOTP_NFSROOT behavior of panicking on a failure to mount the root path provided by the server, unless you've provided an alternative via the ROOTDEVNAME kernel option or by setting vfs.root.mountfrom. The behavior of panicking when given no other options is preserved because it amounts to a bit of a retry loop that could eventually recover from a transient network or server problem. The user can now override the root path from loader(8) even if the kernel is compiled with BOOTP_NFSROOT. If vfs.root.mountfrom is set in the environment it is used unconditionally -- it always overrides the BOOTP info. If it begins with [old]nfs: then the BOOTP code uses it instead of the server-provided info. If it specifies some other filesystem then the bootp code will not panic like it used to and the code in vfs_mountroot.c will invoke the right filesystem to do the mount. If the kernel is compiled with the ROOTDEVNAME option, then that name is used by the BOOTP code if either * The server doesn't provide a pathname. * The boothowto flags include RB_DFLTROOT. The latter allows the user to compile in alternate path in ROOTDEVNAME such as ufs:/dev/da0s1a and boot from that path by setting boot_dftlroot=1 in loader(8) or using the '-r' option in boot(8). The one thing not provided here is automatic failover from a server-provided path to a compiled-in one without the user manually requesting that. The code just isn't currently structured in a way that makes that possible with a lot of rewrite. I think the ability to set vfs.root.mountfrom and to use ROOTDEVNAME automatically when the server doesn't provide a name covers the most common needs. A set of patches submitted by Lars Eggert provided the part I couldn't figure out by myself when I tried to do this last year; many thanks. Reviewed by: rodrigc
2013-07-31 19:14:00 +00:00
if (ifctx->gotrootpath ||
gctx->any_root_overrides) {
gotrootpath = 1;
rtimo = time_second +
BOOTP_SETTLE_DELAY;
Changes to allow using BOOTP_NFSROOT and mounting an nfs root filesystem other than the one specified by the BOOTP server. This configures NFS using the BOOTP protocol while also respecting other root-path options such as setting vfs.root.mountfrom in the environment or using the RB_DFLTROOT boot option. It allows you to override the root path provided by the server, or to supply a root path when the server provides IP configuration but no root path info. This maintains the historical BOOTP_NFSROOT behavior of panicking on a failure to mount the root path provided by the server, unless you've provided an alternative via the ROOTDEVNAME kernel option or by setting vfs.root.mountfrom. The behavior of panicking when given no other options is preserved because it amounts to a bit of a retry loop that could eventually recover from a transient network or server problem. The user can now override the root path from loader(8) even if the kernel is compiled with BOOTP_NFSROOT. If vfs.root.mountfrom is set in the environment it is used unconditionally -- it always overrides the BOOTP info. If it begins with [old]nfs: then the BOOTP code uses it instead of the server-provided info. If it specifies some other filesystem then the bootp code will not panic like it used to and the code in vfs_mountroot.c will invoke the right filesystem to do the mount. If the kernel is compiled with the ROOTDEVNAME option, then that name is used by the BOOTP code if either * The server doesn't provide a pathname. * The boothowto flags include RB_DFLTROOT. The latter allows the user to compile in alternate path in ROOTDEVNAME such as ufs:/dev/da0s1a and boot from that path by setting boot_dftlroot=1 in loader(8) or using the '-r' option in boot(8). The one thing not provided here is automatic failover from a server-provided path to a compiled-in one without the user manually requesting that. The code just isn't currently structured in a way that makes that possible with a lot of rewrite. I think the ability to set vfs.root.mountfrom and to use ROOTDEVNAME automatically when the server doesn't provide a name covers the most common needs. A set of patches submitted by Lars Eggert provided the part I couldn't figure out by myself when I tried to do this last year; many thanks. Reviewed by: rodrigc
2013-07-31 19:14:00 +00:00
if (ifctx->gotrootpath)
printf(" (got root path)");
}
printf("\n");
}
} /* while secs */
#ifdef BOOTP_TIMEOUT
if (gctx->secs > BOOTP_TIMEOUT && BOOTP_TIMEOUT > 0)
break;
#endif
/* Force a retry if halfway in DHCP negotiation */
retry = 0;
STAILQ_FOREACH(ifctx, &gctx->interfaces, next)
if (ifctx->state == IF_DHCP_OFFERED) {
if (ifctx->dhcpquerytype == DHCP_DISCOVER)
retry = 1;
else
ifctx->state = IF_DHCP_UNRESOLVED;
}
if (retry != 0)
continue;
if (gotrootpath != 0) {
gctx->gotrootpath = gotrootpath;
if (rtimo != 0 && time_second >= rtimo)
break;
}
} /* forever send/receive */
/*
* XXX: These are errors of varying seriousness being silently
* ignored
*/
STAILQ_FOREACH(ifctx, &gctx->interfaces, next)
if (bootpc_ifctx_isresolved(ifctx) == 0) {
printf("%s timeout for interface %s\n",
ifctx->dhcpquerytype != DHCP_NOMSG ?
"DHCP" : "BOOTP",
ifctx->ireq.ifr_name);
}
if (gctx->gotrootpath != 0) {
#if 0
printf("Got a root path, ignoring remaining timeout\n");
#endif
error = 0;
goto out;
}
#ifndef BOOTP_NFSROOT
STAILQ_FOREACH(ifctx, &gctx->interfaces, next)
if (bootpc_ifctx_isresolved(ifctx) != 0) {
error = 0;
goto out;
}
#endif
error = ETIMEDOUT;
2000-10-24 22:40:18 +00:00
out:
return (error);
}
static void
bootpc_fakeup_interface(struct bootpc_ifcontext *ifctx, struct thread *td)
{
struct ifreq *ifr;
struct in_aliasreq *ifra;
2000-10-24 22:40:18 +00:00
struct sockaddr_in *sin;
int error;
ifr = &ifctx->ireq;
ifra = &ifctx->iareq;
2000-10-24 22:40:18 +00:00
/*
* Bring up the interface.
*
* Get the old interface flags and or IFF_UP into them; if
* IFF_UP set blindly, interface selection can be clobbered.
*/
error = ifioctl(bootp_so, SIOCGIFFLAGS, (caddr_t)ifr, td);
if (error != 0)
panic("%s: SIOCGIFFLAGS, error=%d", __func__, error);
ifr->ifr_flags |= IFF_UP;
error = ifioctl(bootp_so, SIOCSIFFLAGS, (caddr_t)ifr, td);
if (error != 0)
panic("%s: SIOCSIFFLAGS, error=%d", __func__, error);
2000-10-24 22:40:18 +00:00
/*
* Do enough of ifconfig(8) so that the chosen interface
* can talk to the servers. Set address to 0.0.0.0/8 and
* broadcast address to local broadcast.
2000-10-24 22:40:18 +00:00
*/
sin = (struct sockaddr_in *)&ifra->ifra_addr;
clear_sinaddr(sin);
sin = (struct sockaddr_in *)&ifra->ifra_mask;
clear_sinaddr(sin);
sin->sin_addr.s_addr = htonl(IN_CLASSA_NET);
sin = (struct sockaddr_in *)&ifra->ifra_broadaddr;
clear_sinaddr(sin);
sin->sin_addr.s_addr = htonl(INADDR_BROADCAST);
error = ifioctl(bootp_so, SIOCAIFADDR, (caddr_t)ifra, td);
if (error != 0)
panic("%s: SIOCAIFADDR, error=%d", __func__, error);
}
static void
bootpc_shutdown_interface(struct bootpc_ifcontext *ifctx, struct thread *td)
{
struct ifreq *ifr;
struct sockaddr_in *sin;
int error;
ifr = &ifctx->ireq;
printf("Shutdown interface %s\n", ifctx->ireq.ifr_name);
error = ifioctl(bootp_so, SIOCGIFFLAGS, (caddr_t)ifr, td);
if (error != 0)
panic("%s: SIOCGIFFLAGS, error=%d", __func__, error);
ifr->ifr_flags &= ~IFF_UP;
error = ifioctl(bootp_so, SIOCSIFFLAGS, (caddr_t)ifr, td);
if (error != 0)
panic("%s: SIOCSIFFLAGS, error=%d", __func__, error);
sin = (struct sockaddr_in *) &ifr->ifr_addr;
clear_sinaddr(sin);
error = ifioctl(bootp_so, SIOCDIFADDR, (caddr_t) ifr, td);
if (error != 0)
panic("%s: SIOCDIFADDR, error=%d", __func__, error);
}
static void
bootpc_adjust_interface(struct bootpc_ifcontext *ifctx,
struct bootpc_globalcontext *gctx, struct thread *td)
{
2000-10-24 22:40:18 +00:00
int error;
struct sockaddr_in *sin;
struct ifreq *ifr;
struct in_aliasreq *ifra;
struct sockaddr_in *myaddr;
struct sockaddr_in *netmask;
ifr = &ifctx->ireq;
ifra = &ifctx->iareq;
myaddr = &ifctx->myaddr;
netmask = &ifctx->netmask;
if (bootpc_ifctx_isresolved(ifctx) == 0) {
/* Shutdown interfaces where BOOTP failed */
bootpc_shutdown_interface(ifctx, td);
return;
2000-10-24 22:40:18 +00:00
}
printf("Adjusted interface %s", ifctx->ireq.ifr_name);
/* Do BOOTP interface options */
if (ifctx->mtu != 0) {
printf(" (MTU=%d%s)", ifctx->mtu,
(ifctx->mtu > 1514) ? "/JUMBO" : "");
ifr->ifr_mtu = ifctx->mtu;
error = ifioctl(bootp_so, SIOCSIFMTU, (caddr_t) ifr, td);
if (error != 0)
panic("%s: SIOCSIFMTU, error=%d", __func__, error);
}
printf("\n");
2000-10-24 22:40:18 +00:00
/*
* Do enough of ifconfig(8) so that the chosen interface
* can talk to the servers. (just set the address)
*/
sin = (struct sockaddr_in *) &ifr->ifr_addr;
clear_sinaddr(sin);
error = ifioctl(bootp_so, SIOCDIFADDR, (caddr_t) ifr, td);
2000-10-24 22:40:18 +00:00
if (error != 0)
panic("%s: SIOCDIFADDR, error=%d", __func__, error);
bcopy(myaddr, &ifra->ifra_addr, sizeof(*myaddr));
bcopy(netmask, &ifra->ifra_mask, sizeof(*netmask));
clear_sinaddr(&ifra->ifra_broadaddr);
ifra->ifra_broadaddr.sin_addr.s_addr = myaddr->sin_addr.s_addr |
~netmask->sin_addr.s_addr;
error = ifioctl(bootp_so, SIOCAIFADDR, (caddr_t)ifra, td);
2000-10-24 22:40:18 +00:00
if (error != 0)
panic("%s: SIOCAIFADDR, error=%d", __func__, error);
}
static void
bootpc_add_default_route(struct bootpc_ifcontext *ifctx)
{
int error;
struct sockaddr_in defdst;
struct sockaddr_in defmask;
struct rt_addrinfo info;
struct rib_cmd_info rc;
if (ifctx->gw.sin_addr.s_addr == htonl(INADDR_ANY))
return;
2000-10-24 22:40:18 +00:00
clear_sinaddr(&defdst);
clear_sinaddr(&defmask);
bzero((caddr_t)&info, sizeof(info));
info.rti_flags = RTF_UP | RTF_GATEWAY | RTF_STATIC;
info.rti_info[RTAX_DST] = (struct sockaddr *)&defdst;
info.rti_info[RTAX_NETMASK] = (struct sockaddr *)&defmask;
info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ifctx->gw;
error = rib_action(RT_DEFAULT_FIB, RTM_ADD, &info, &rc);
if (error != 0) {
printf("%s: RTM_ADD, error=%d\n", __func__, error);
}
}
static void
bootpc_remove_default_route(struct bootpc_ifcontext *ifctx)
{
int error;
struct sockaddr_in defdst;
struct sockaddr_in defmask;
struct rt_addrinfo info;
struct rib_cmd_info rc;
if (ifctx->gw.sin_addr.s_addr == htonl(INADDR_ANY))
return;
clear_sinaddr(&defdst);
clear_sinaddr(&defmask);
bzero((caddr_t)&info, sizeof(info));
info.rti_flags = RTF_UP | RTF_GATEWAY | RTF_STATIC;
info.rti_info[RTAX_DST] = (struct sockaddr *)&defdst;
info.rti_info[RTAX_NETMASK] = (struct sockaddr *)&defmask;
info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&ifctx->gw;
error = rib_action(RT_DEFAULT_FIB, RTM_DELETE, &info, &rc);
if (error != 0) {
printf("%s: RTM_DELETE, error=%d\n", __func__, error);
2000-10-24 22:40:18 +00:00
}
}
2000-10-24 22:40:18 +00:00
static int
setfs(struct sockaddr_in *addr, char *path, char *p,
const struct in_addr *siaddr)
{
if (getip(&p, &addr->sin_addr) == 0) {
if (siaddr != NULL && *p == '/')
bcopy(siaddr, &addr->sin_addr, sizeof(struct in_addr));
else
return 0;
} else {
if (*p != ':')
return 0;
p++;
}
addr->sin_len = sizeof(struct sockaddr_in);
addr->sin_family = AF_INET;
strlcpy(path, p, MNAMELEN);
return 1;
}
static int
getip(char **ptr, struct in_addr *addr)
{
char *p;
2000-10-24 22:40:18 +00:00
unsigned int ip;
int val;
p = *ptr;
2000-10-24 22:40:18 +00:00
ip = 0;
if (((val = getdec(&p)) < 0) || (val > 255))
return 0;
ip = val << 24;
2000-10-24 22:40:18 +00:00
if (*p != '.')
return 0;
p++;
2000-10-24 22:40:18 +00:00
if (((val = getdec(&p)) < 0) || (val > 255))
return 0;
ip |= (val << 16);
2000-10-24 22:40:18 +00:00
if (*p != '.')
return 0;
p++;
2000-10-24 22:40:18 +00:00
if (((val = getdec(&p)) < 0) || (val > 255))
return 0;
ip |= (val << 8);
2000-10-24 22:40:18 +00:00
if (*p != '.')
return 0;
p++;
2000-10-24 22:40:18 +00:00
if (((val = getdec(&p)) < 0) || (val > 255))
return 0;
ip |= val;
addr->s_addr = htonl(ip);
*ptr = p;
2000-10-24 22:40:18 +00:00
return 1;
}
2000-10-24 22:40:18 +00:00
static int
getdec(char **ptr)
{
2000-10-24 22:40:18 +00:00
char *p;
int ret;
p = *ptr;
ret = 0;
if ((*p < '0') || (*p > '9'))
return -1;
while ((*p >= '0') && (*p <= '9')) {
2000-10-24 22:40:18 +00:00
ret = ret * 10 + (*p - '0');
p++;
}
*ptr = p;
2000-10-24 22:40:18 +00:00
return ret;
}
2000-10-24 22:40:18 +00:00
static void
mountopts(struct nfs_args *args, char *p)
{
args->version = NFS_ARGSVERSION;
args->rsize = BOOTP_BLOCKSIZE;
args->wsize = BOOTP_BLOCKSIZE;
args->flags = NFSMNT_RSIZE | NFSMNT_WSIZE | NFSMNT_RESVPORT;
args->sotype = SOCK_DGRAM;
if (p != NULL)
nfs_parse_options(p, args);
}
2000-10-24 22:40:18 +00:00
static int
xdr_opaque_decode(struct mbuf **mptr, u_char *buf, int len)
{
2000-10-24 22:40:18 +00:00
struct mbuf *m;
int alignedlen;
2000-10-24 22:40:18 +00:00
m = *mptr;
alignedlen = ( len + 3 ) & ~3;
2000-10-24 22:40:18 +00:00
if (m->m_len < alignedlen) {
m = m_pullup(m, alignedlen);
if (m == NULL) {
*mptr = NULL;
return EBADRPC;
}
}
bcopy(mtod(m, u_char *), buf, len);
m_adj(m, alignedlen);
*mptr = m;
return 0;
}
2000-10-24 22:40:18 +00:00
static int
xdr_int_decode(struct mbuf **mptr, int *iptr)
{
2000-10-24 22:40:18 +00:00
u_int32_t i;
2000-10-24 22:40:18 +00:00
if (xdr_opaque_decode(mptr, (u_char *) &i, sizeof(u_int32_t)) != 0)
return EBADRPC;
*iptr = fxdr_unsigned(u_int32_t, i);
return 0;
}
2000-10-24 22:40:18 +00:00
static void
print_sin_addr(struct sockaddr_in *sin)
{
print_in_addr(sin->sin_addr);
}
static void
print_in_addr(struct in_addr addr)
{
unsigned int ip;
ip = ntohl(addr.s_addr);
printf("%d.%d.%d.%d",
2000-10-24 22:40:18 +00:00
ip >> 24, (ip >> 16) & 255, (ip >> 8) & 255, ip & 255);
}
static void
bootpc_compose_query(struct bootpc_ifcontext *ifctx, struct thread *td)
{
unsigned char *vendp;
unsigned char vendor_client[64];
uint32_t leasetime;
uint8_t vendor_client_len;
ifctx->gotrootpath = 0;
bzero((caddr_t) &ifctx->call, sizeof(ifctx->call));
/* bootpc part */
ifctx->call.op = BOOTP_REQUEST; /* BOOTREQUEST */
ifctx->call.htype = 1; /* 10mb ethernet */
ifctx->call.hlen = ifctx->sdl->sdl_alen;/* Hardware address length */
ifctx->call.hops = 0;
if (bootpc_ifctx_isunresolved(ifctx) != 0)
ifctx->xid++;
ifctx->call.xid = txdr_unsigned(ifctx->xid);
bcopy(LLADDR(ifctx->sdl), &ifctx->call.chaddr, ifctx->sdl->sdl_alen);
vendp = ifctx->call.vend;
*vendp++ = 99; /* RFC1048 cookie */
*vendp++ = 130;
*vendp++ = 83;
*vendp++ = 99;
*vendp++ = TAG_MAXMSGSIZE;
*vendp++ = 2;
*vendp++ = (sizeof(struct bootp_packet) >> 8) & 255;
*vendp++ = sizeof(struct bootp_packet) & 255;
snprintf(vendor_client, sizeof(vendor_client), "%s:%s:%s",
ostype, MACHINE, osrelease);
vendor_client_len = strlen(vendor_client);
*vendp++ = TAG_VENDOR_INDENTIFIER;
*vendp++ = vendor_client_len;
memcpy(vendp, vendor_client, vendor_client_len);
vendp += vendor_client_len;
ifctx->dhcpquerytype = DHCP_NOMSG;
switch (ifctx->state) {
case IF_DHCP_UNRESOLVED:
*vendp++ = TAG_DHCP_MSGTYPE;
*vendp++ = 1;
*vendp++ = DHCP_DISCOVER;
ifctx->dhcpquerytype = DHCP_DISCOVER;
ifctx->gotdhcpserver = 0;
break;
case IF_DHCP_OFFERED:
*vendp++ = TAG_DHCP_MSGTYPE;
*vendp++ = 1;
*vendp++ = DHCP_REQUEST;
ifctx->dhcpquerytype = DHCP_REQUEST;
*vendp++ = TAG_DHCP_REQ_ADDR;
*vendp++ = 4;
memcpy(vendp, &ifctx->reply.yiaddr, 4);
vendp += 4;
if (ifctx->gotdhcpserver != 0) {
*vendp++ = TAG_DHCP_SERVERID;
*vendp++ = 4;
memcpy(vendp, &ifctx->dhcpserver, 4);
vendp += 4;
}
*vendp++ = TAG_DHCP_LEASETIME;
*vendp++ = 4;
leasetime = htonl(300);
memcpy(vendp, &leasetime, 4);
vendp += 4;
break;
default:
break;
}
*vendp = TAG_END;
ifctx->call.secs = 0;
ifctx->call.flags = htons(0x8000); /* We need a broadcast answer */
}
2000-10-24 22:40:18 +00:00
static int
bootpc_hascookie(struct bootp_packet *bp)
{
return (bp->vend[0] == 99 && bp->vend[1] == 130 &&
bp->vend[2] == 83 && bp->vend[3] == 99);
}
static void
bootpc_tag_helper(struct bootpc_tagcontext *tctx,
unsigned char *start, int len, int tag)
{
unsigned char *j;
unsigned char *ej;
unsigned char code;
if (tctx->badtag != 0 || tctx->badopt != 0)
return;
j = start;
ej = j + len;
while (j < ej) {
code = *j++;
if (code == TAG_PAD)
continue;
if (code == TAG_END)
return;
if (j >= ej || j + *j + 1 > ej) {
tctx->badopt = 1;
return;
}
len = *j++;
if (code == tag) {
if (tctx->taglen + len > TAG_MAXLEN) {
tctx->badtag = 1;
return;
}
tctx->foundopt = 1;
if (len > 0)
memcpy(tctx->buf + tctx->taglen,
j, len);
tctx->taglen += len;
}
if (code == TAG_OVERLOAD)
tctx->overload = *j;
j += len;
}
}
static unsigned char *
bootpc_tag(struct bootpc_tagcontext *tctx,
struct bootp_packet *bp, int len, int tag)
{
tctx->overload = 0;
tctx->badopt = 0;
tctx->badtag = 0;
tctx->foundopt = 0;
tctx->taglen = 0;
if (bootpc_hascookie(bp) == 0)
return NULL;
bootpc_tag_helper(tctx, &bp->vend[4],
(unsigned char *) bp + len - &bp->vend[4], tag);
if ((tctx->overload & OVERLOAD_FILE) != 0)
bootpc_tag_helper(tctx,
(unsigned char *) bp->file,
sizeof(bp->file),
tag);
if ((tctx->overload & OVERLOAD_SNAME) != 0)
bootpc_tag_helper(tctx,
(unsigned char *) bp->sname,
sizeof(bp->sname),
tag);
if (tctx->badopt != 0 || tctx->badtag != 0 || tctx->foundopt == 0)
return NULL;
tctx->buf[tctx->taglen] = '\0';
return tctx->buf;
}
static void
bootpc_decode_reply(struct nfsv3_diskless *nd, struct bootpc_ifcontext *ifctx,
struct bootpc_globalcontext *gctx)
{
Changes to allow using BOOTP_NFSROOT and mounting an nfs root filesystem other than the one specified by the BOOTP server. This configures NFS using the BOOTP protocol while also respecting other root-path options such as setting vfs.root.mountfrom in the environment or using the RB_DFLTROOT boot option. It allows you to override the root path provided by the server, or to supply a root path when the server provides IP configuration but no root path info. This maintains the historical BOOTP_NFSROOT behavior of panicking on a failure to mount the root path provided by the server, unless you've provided an alternative via the ROOTDEVNAME kernel option or by setting vfs.root.mountfrom. The behavior of panicking when given no other options is preserved because it amounts to a bit of a retry loop that could eventually recover from a transient network or server problem. The user can now override the root path from loader(8) even if the kernel is compiled with BOOTP_NFSROOT. If vfs.root.mountfrom is set in the environment it is used unconditionally -- it always overrides the BOOTP info. If it begins with [old]nfs: then the BOOTP code uses it instead of the server-provided info. If it specifies some other filesystem then the bootp code will not panic like it used to and the code in vfs_mountroot.c will invoke the right filesystem to do the mount. If the kernel is compiled with the ROOTDEVNAME option, then that name is used by the BOOTP code if either * The server doesn't provide a pathname. * The boothowto flags include RB_DFLTROOT. The latter allows the user to compile in alternate path in ROOTDEVNAME such as ufs:/dev/da0s1a and boot from that path by setting boot_dftlroot=1 in loader(8) or using the '-r' option in boot(8). The one thing not provided here is automatic failover from a server-provided path to a compiled-in one without the user manually requesting that. The code just isn't currently structured in a way that makes that possible with a lot of rewrite. I think the ability to set vfs.root.mountfrom and to use ROOTDEVNAME automatically when the server doesn't provide a name covers the most common needs. A set of patches submitted by Lars Eggert provided the part I couldn't figure out by myself when I tried to do this last year; many thanks. Reviewed by: rodrigc
2013-07-31 19:14:00 +00:00
char *p, *s;
2000-10-24 22:40:18 +00:00
unsigned int ip;
ifctx->gotgw = 0;
ifctx->gotnetmask = 0;
clear_sinaddr(&ifctx->myaddr);
clear_sinaddr(&ifctx->netmask);
clear_sinaddr(&ifctx->gw);
ifctx->myaddr.sin_addr = ifctx->reply.yiaddr;
ip = ntohl(ifctx->myaddr.sin_addr.s_addr);
printf("%s at ", ifctx->ireq.ifr_name);
print_sin_addr(&ifctx->myaddr);
printf(" server ");
print_in_addr(ifctx->reply.siaddr);
ifctx->gw.sin_addr = ifctx->reply.giaddr;
if (ifctx->reply.giaddr.s_addr != htonl(INADDR_ANY)) {
printf(" via gateway ");
print_in_addr(ifctx->reply.giaddr);
}
/* This call used for the side effect (overload flag) */
(void) bootpc_tag(&gctx->tmptag,
&ifctx->reply, ifctx->replylen, TAG_END);
if ((gctx->tmptag.overload & OVERLOAD_SNAME) == 0)
if (ifctx->reply.sname[0] != '\0')
printf(" server name %s", ifctx->reply.sname);
if ((gctx->tmptag.overload & OVERLOAD_FILE) == 0)
if (ifctx->reply.file[0] != '\0')
printf(" boot file %s", ifctx->reply.file);
printf("\n");
p = bootpc_tag(&gctx->tag, &ifctx->reply, ifctx->replylen,
TAG_SUBNETMASK);
if (p != NULL) {
if (gctx->tag.taglen != 4)
panic("bootpc: subnet mask len is %d",
gctx->tag.taglen);
bcopy(p, &ifctx->netmask.sin_addr, 4);
ifctx->gotnetmask = 1;
printf("subnet mask ");
print_sin_addr(&ifctx->netmask);
printf(" ");
}
p = bootpc_tag(&gctx->tag, &ifctx->reply, ifctx->replylen,
TAG_ROUTERS);
if (p != NULL) {
/* Routers */
if (gctx->tag.taglen % 4)
panic("bootpc: Router Len is %d", gctx->tag.taglen);
if (gctx->tag.taglen > 0) {
bcopy(p, &ifctx->gw.sin_addr, 4);
printf("router ");
print_sin_addr(&ifctx->gw);
printf(" ");
ifctx->gotgw = 1;
gctx->gotgw = 1;
}
}
Changes to allow using BOOTP_NFSROOT and mounting an nfs root filesystem other than the one specified by the BOOTP server. This configures NFS using the BOOTP protocol while also respecting other root-path options such as setting vfs.root.mountfrom in the environment or using the RB_DFLTROOT boot option. It allows you to override the root path provided by the server, or to supply a root path when the server provides IP configuration but no root path info. This maintains the historical BOOTP_NFSROOT behavior of panicking on a failure to mount the root path provided by the server, unless you've provided an alternative via the ROOTDEVNAME kernel option or by setting vfs.root.mountfrom. The behavior of panicking when given no other options is preserved because it amounts to a bit of a retry loop that could eventually recover from a transient network or server problem. The user can now override the root path from loader(8) even if the kernel is compiled with BOOTP_NFSROOT. If vfs.root.mountfrom is set in the environment it is used unconditionally -- it always overrides the BOOTP info. If it begins with [old]nfs: then the BOOTP code uses it instead of the server-provided info. If it specifies some other filesystem then the bootp code will not panic like it used to and the code in vfs_mountroot.c will invoke the right filesystem to do the mount. If the kernel is compiled with the ROOTDEVNAME option, then that name is used by the BOOTP code if either * The server doesn't provide a pathname. * The boothowto flags include RB_DFLTROOT. The latter allows the user to compile in alternate path in ROOTDEVNAME such as ufs:/dev/da0s1a and boot from that path by setting boot_dftlroot=1 in loader(8) or using the '-r' option in boot(8). The one thing not provided here is automatic failover from a server-provided path to a compiled-in one without the user manually requesting that. The code just isn't currently structured in a way that makes that possible with a lot of rewrite. I think the ability to set vfs.root.mountfrom and to use ROOTDEVNAME automatically when the server doesn't provide a name covers the most common needs. A set of patches submitted by Lars Eggert provided the part I couldn't figure out by myself when I tried to do this last year; many thanks. Reviewed by: rodrigc
2013-07-31 19:14:00 +00:00
/*
* Choose a root filesystem. If a value is forced in the environment
* and it contains "nfs:", use it unconditionally. Otherwise, if the
* kernel is compiled with the ROOTDEVNAME option, then use it if:
* - The server doesn't provide a pathname.
* - The boothowto flags include RB_DFLTROOT (user said to override
* the server value).
*/
p = NULL;
if ((s = kern_getenv("vfs.root.mountfrom")) != NULL) {
Changes to allow using BOOTP_NFSROOT and mounting an nfs root filesystem other than the one specified by the BOOTP server. This configures NFS using the BOOTP protocol while also respecting other root-path options such as setting vfs.root.mountfrom in the environment or using the RB_DFLTROOT boot option. It allows you to override the root path provided by the server, or to supply a root path when the server provides IP configuration but no root path info. This maintains the historical BOOTP_NFSROOT behavior of panicking on a failure to mount the root path provided by the server, unless you've provided an alternative via the ROOTDEVNAME kernel option or by setting vfs.root.mountfrom. The behavior of panicking when given no other options is preserved because it amounts to a bit of a retry loop that could eventually recover from a transient network or server problem. The user can now override the root path from loader(8) even if the kernel is compiled with BOOTP_NFSROOT. If vfs.root.mountfrom is set in the environment it is used unconditionally -- it always overrides the BOOTP info. If it begins with [old]nfs: then the BOOTP code uses it instead of the server-provided info. If it specifies some other filesystem then the bootp code will not panic like it used to and the code in vfs_mountroot.c will invoke the right filesystem to do the mount. If the kernel is compiled with the ROOTDEVNAME option, then that name is used by the BOOTP code if either * The server doesn't provide a pathname. * The boothowto flags include RB_DFLTROOT. The latter allows the user to compile in alternate path in ROOTDEVNAME such as ufs:/dev/da0s1a and boot from that path by setting boot_dftlroot=1 in loader(8) or using the '-r' option in boot(8). The one thing not provided here is automatic failover from a server-provided path to a compiled-in one without the user manually requesting that. The code just isn't currently structured in a way that makes that possible with a lot of rewrite. I think the ability to set vfs.root.mountfrom and to use ROOTDEVNAME automatically when the server doesn't provide a name covers the most common needs. A set of patches submitted by Lars Eggert provided the part I couldn't figure out by myself when I tried to do this last year; many thanks. Reviewed by: rodrigc
2013-07-31 19:14:00 +00:00
if ((p = strstr(s, "nfs:")) != NULL)
p = strdup(p + 4, M_TEMP);
freeenv(s);
}
if (p == NULL) {
p = bootpc_tag(&gctx->tag, &ifctx->reply, ifctx->replylen,
TAG_ROOT);
if (p != NULL)
ifctx->gotrootpath = 1;
Changes to allow using BOOTP_NFSROOT and mounting an nfs root filesystem other than the one specified by the BOOTP server. This configures NFS using the BOOTP protocol while also respecting other root-path options such as setting vfs.root.mountfrom in the environment or using the RB_DFLTROOT boot option. It allows you to override the root path provided by the server, or to supply a root path when the server provides IP configuration but no root path info. This maintains the historical BOOTP_NFSROOT behavior of panicking on a failure to mount the root path provided by the server, unless you've provided an alternative via the ROOTDEVNAME kernel option or by setting vfs.root.mountfrom. The behavior of panicking when given no other options is preserved because it amounts to a bit of a retry loop that could eventually recover from a transient network or server problem. The user can now override the root path from loader(8) even if the kernel is compiled with BOOTP_NFSROOT. If vfs.root.mountfrom is set in the environment it is used unconditionally -- it always overrides the BOOTP info. If it begins with [old]nfs: then the BOOTP code uses it instead of the server-provided info. If it specifies some other filesystem then the bootp code will not panic like it used to and the code in vfs_mountroot.c will invoke the right filesystem to do the mount. If the kernel is compiled with the ROOTDEVNAME option, then that name is used by the BOOTP code if either * The server doesn't provide a pathname. * The boothowto flags include RB_DFLTROOT. The latter allows the user to compile in alternate path in ROOTDEVNAME such as ufs:/dev/da0s1a and boot from that path by setting boot_dftlroot=1 in loader(8) or using the '-r' option in boot(8). The one thing not provided here is automatic failover from a server-provided path to a compiled-in one without the user manually requesting that. The code just isn't currently structured in a way that makes that possible with a lot of rewrite. I think the ability to set vfs.root.mountfrom and to use ROOTDEVNAME automatically when the server doesn't provide a name covers the most common needs. A set of patches submitted by Lars Eggert provided the part I couldn't figure out by myself when I tried to do this last year; many thanks. Reviewed by: rodrigc
2013-07-31 19:14:00 +00:00
}
#ifdef ROOTDEVNAME
if ((p == NULL || (boothowto & RB_DFLTROOT) != 0) &&
(p = strstr(ROOTDEVNAME, "nfs:")) != NULL) {
p += 4;
}
#endif
if (p != NULL) {
if (gctx->setrootfs != NULL) {
printf("rootfs %s (ignored) ", p);
} else if (setfs(&nd->root_saddr,
nd->root_hostnam, p, &ifctx->reply.siaddr)) {
if (*p == '/') {
printf("root_server ");
print_sin_addr(&nd->root_saddr);
printf(" ");
}
printf("rootfs %s ", p);
gctx->gotrootpath = 1;
gctx->setrootfs = ifctx;
p = bootpc_tag(&gctx->tag, &ifctx->reply,
ifctx->replylen,
TAG_ROOTOPTS);
if (p != NULL) {
mountopts(&nd->root_args, p);
printf("rootopts %s ", p);
}
} else
panic("Failed to set rootfs to %s", p);
}
p = bootpc_tag(&gctx->tag, &ifctx->reply, ifctx->replylen,
TAG_HOSTNAME);
if (p != NULL) {
if (gctx->tag.taglen >= MAXHOSTNAMELEN)
panic("bootpc: hostname >= %d bytes",
MAXHOSTNAMELEN);
if (gctx->sethostname != NULL) {
printf("hostname %s (ignored) ", p);
} else {
strcpy(nd->my_hostnam, p);
mtx_lock(&prison0.pr_mtx);
strcpy(prison0.pr_hostname, p);
mtx_unlock(&prison0.pr_mtx);
printf("hostname %s ", p);
gctx->sethostname = ifctx;
}
}
p = bootpc_tag(&gctx->tag, &ifctx->reply, ifctx->replylen,
TAG_COOKIE);
if (p != NULL) { /* store in a sysctl variable */
int i, l = sizeof(bootp_cookie) - 1;
for (i = 0; i < l && p[i] != '\0'; i++)
bootp_cookie[i] = p[i];
p[i] = '\0';
}
p = bootpc_tag(&gctx->tag, &ifctx->reply, ifctx->replylen,
TAG_INTF_MTU);
if (p != NULL) {
ifctx->mtu = be16dec(p);
}
printf("\n");
if (ifctx->gotnetmask == 0) {
if (IN_CLASSA(ntohl(ifctx->myaddr.sin_addr.s_addr)))
ifctx->netmask.sin_addr.s_addr = htonl(IN_CLASSA_NET);
else if (IN_CLASSB(ntohl(ifctx->myaddr.sin_addr.s_addr)))
ifctx->netmask.sin_addr.s_addr = htonl(IN_CLASSB_NET);
else
ifctx->netmask.sin_addr.s_addr = htonl(IN_CLASSC_NET);
}
}
2000-10-24 22:40:18 +00:00
void
bootpc_init(void)
{
struct bootpc_ifcontext *ifctx; /* Interface BOOTP contexts */
struct bootpc_globalcontext *gctx; /* Global BOOTP context */
struct ifnet *ifp;
struct sockaddr_dl *sdl;
struct ifaddr *ifa;
int error;
#ifndef BOOTP_WIRED_TO
int ifcnt;
#endif
2000-10-24 22:40:18 +00:00
struct nfsv3_diskless *nd;
struct thread *td;
int timeout;
int delay;
timeout = BOOTP_IFACE_WAIT_TIMEOUT * hz;
delay = hz / 10;
2000-10-24 22:40:18 +00:00
nd = &nfsv3_diskless;
td = curthread;
2000-10-24 22:40:18 +00:00
/*
* If already filled in, don't touch it here
*/
if (nfs_diskless_valid != 0)
return;
gctx = malloc(sizeof(*gctx), M_TEMP, M_WAITOK | M_ZERO);
STAILQ_INIT(&gctx->interfaces);
gctx->xid = ~0xFFFF;
gctx->starttime = time_second;
Changes to allow using BOOTP_NFSROOT and mounting an nfs root filesystem other than the one specified by the BOOTP server. This configures NFS using the BOOTP protocol while also respecting other root-path options such as setting vfs.root.mountfrom in the environment or using the RB_DFLTROOT boot option. It allows you to override the root path provided by the server, or to supply a root path when the server provides IP configuration but no root path info. This maintains the historical BOOTP_NFSROOT behavior of panicking on a failure to mount the root path provided by the server, unless you've provided an alternative via the ROOTDEVNAME kernel option or by setting vfs.root.mountfrom. The behavior of panicking when given no other options is preserved because it amounts to a bit of a retry loop that could eventually recover from a transient network or server problem. The user can now override the root path from loader(8) even if the kernel is compiled with BOOTP_NFSROOT. If vfs.root.mountfrom is set in the environment it is used unconditionally -- it always overrides the BOOTP info. If it begins with [old]nfs: then the BOOTP code uses it instead of the server-provided info. If it specifies some other filesystem then the bootp code will not panic like it used to and the code in vfs_mountroot.c will invoke the right filesystem to do the mount. If the kernel is compiled with the ROOTDEVNAME option, then that name is used by the BOOTP code if either * The server doesn't provide a pathname. * The boothowto flags include RB_DFLTROOT. The latter allows the user to compile in alternate path in ROOTDEVNAME such as ufs:/dev/da0s1a and boot from that path by setting boot_dftlroot=1 in loader(8) or using the '-r' option in boot(8). The one thing not provided here is automatic failover from a server-provided path to a compiled-in one without the user manually requesting that. The code just isn't currently structured in a way that makes that possible with a lot of rewrite. I think the ability to set vfs.root.mountfrom and to use ROOTDEVNAME automatically when the server doesn't provide a name covers the most common needs. A set of patches submitted by Lars Eggert provided the part I couldn't figure out by myself when I tried to do this last year; many thanks. Reviewed by: rodrigc
2013-07-31 19:14:00 +00:00
/*
* If ROOTDEVNAME is defined or vfs.root.mountfrom is set then we have
* root-path overrides that can potentially let us boot even if we don't
* get a root path from the server, so we can treat that as a non-error.
*/
#ifdef ROOTDEVNAME
gctx->any_root_overrides = 1;
#else
gctx->any_root_overrides = testenv("vfs.root.mountfrom");
#endif
2000-10-24 22:40:18 +00:00
/*
* Find a network interface.
*/
CURVNET_SET(TD_TO_VNET(td));
#ifdef BOOTP_WIRED_TO
printf("%s: wired to interface '%s'\n", __func__,
__XSTRING(BOOTP_WIRED_TO));
allocifctx(gctx);
#else
/*
* Preallocate interface context storage, if another interface
* attaches and wins the race, it won't be eligible for bootp.
*/
ifcnt = 0;
IFNET_RLOCK();
CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
if ((ifp->if_flags &
(IFF_LOOPBACK | IFF_POINTOPOINT | IFF_BROADCAST)) !=
IFF_BROADCAST)
continue;
switch (ifp->if_alloctype) {
case IFT_ETHER:
break;
default:
continue;
}
ifcnt++;
}
IFNET_RUNLOCK();
if (ifcnt == 0)
panic("%s: no eligible interfaces", __func__);
for (; ifcnt > 0; ifcnt--)
allocifctx(gctx);
#endif
retry:
ifctx = STAILQ_FIRST(&gctx->interfaces);
IFNET_RLOCK();
CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
if (ifctx == NULL)
break;
#ifdef BOOTP_WIRED_TO
if (strcmp(ifp->if_xname, __XSTRING(BOOTP_WIRED_TO)) != 0)
continue;
#else
if ((ifp->if_flags &
(IFF_LOOPBACK | IFF_POINTOPOINT | IFF_BROADCAST)) !=
IFF_BROADCAST)
continue;
switch (ifp->if_alloctype) {
case IFT_ETHER:
break;
default:
continue;
}
#endif
strlcpy(ifctx->ireq.ifr_name, ifp->if_xname,
sizeof(ifctx->ireq.ifr_name));
ifctx->ifp = ifp;
/* Get HW address */
sdl = NULL;
ifnet: Replace if_addr_lock rwlock with epoch + mutex Run on LLNW canaries and tested by pho@ gallatin: Using a 14-core, 28-HTT single socket E5-2697 v3 with a 40GbE MLX5 based ConnectX 4-LX NIC, I see an almost 12% improvement in received packet rate, and a larger improvement in bytes delivered all the way to userspace. When the host receiving 64 streams of netperf -H $DUT -t UDP_STREAM -- -m 1, I see, using nstat -I mce0 1 before the patch: InMpps OMpps InGbs OGbs err TCP Est %CPU syscalls csw irq GBfree 4.98 0.00 4.42 0.00 4235592 33 83.80 4720653 2149771 1235 247.32 4.73 0.00 4.20 0.00 4025260 33 82.99 4724900 2139833 1204 247.32 4.72 0.00 4.20 0.00 4035252 33 82.14 4719162 2132023 1264 247.32 4.71 0.00 4.21 0.00 4073206 33 83.68 4744973 2123317 1347 247.32 4.72 0.00 4.21 0.00 4061118 33 80.82 4713615 2188091 1490 247.32 4.72 0.00 4.21 0.00 4051675 33 85.29 4727399 2109011 1205 247.32 4.73 0.00 4.21 0.00 4039056 33 84.65 4724735 2102603 1053 247.32 After the patch InMpps OMpps InGbs OGbs err TCP Est %CPU syscalls csw irq GBfree 5.43 0.00 4.20 0.00 3313143 33 84.96 5434214 1900162 2656 245.51 5.43 0.00 4.20 0.00 3308527 33 85.24 5439695 1809382 2521 245.51 5.42 0.00 4.19 0.00 3316778 33 87.54 5416028 1805835 2256 245.51 5.42 0.00 4.19 0.00 3317673 33 90.44 5426044 1763056 2332 245.51 5.42 0.00 4.19 0.00 3314839 33 88.11 5435732 1792218 2499 245.52 5.44 0.00 4.19 0.00 3293228 33 91.84 5426301 1668597 2121 245.52 Similarly, netperf reports 230Mb/s before the patch, and 270Mb/s after the patch Reviewed by: gallatin Sponsored by: Limelight Networks Differential Revision: https://reviews.freebsd.org/D15366
2018-05-18 20:13:34 +00:00
CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link)
if (ifa->ifa_addr->sa_family == AF_LINK) {
sdl = (struct sockaddr_dl *)ifa->ifa_addr;
if (sdl->sdl_type == IFT_ETHER)
break;
}
if (sdl == NULL)
panic("bootpc: Unable to find HW address for %s",
ifctx->ireq.ifr_name);
ifctx->sdl = sdl;
ifctx = STAILQ_NEXT(ifctx, next);
2000-10-24 22:40:18 +00:00
}
2002-12-22 05:35:03 +00:00
IFNET_RUNLOCK();
CURVNET_RESTORE();
if (STAILQ_EMPTY(&gctx->interfaces) ||
STAILQ_FIRST(&gctx->interfaces)->ifp == NULL) {
if (timeout > 0) {
pause("bootpc", delay);
timeout -= delay;
goto retry;
}
#ifdef BOOTP_WIRED_TO
panic("%s: Could not find interface specified "
"by BOOTP_WIRED_TO: "
__XSTRING(BOOTP_WIRED_TO), __func__);
#else
panic("%s: no suitable interface", __func__);
#endif
}
error = socreate(AF_INET, &bootp_so, SOCK_DGRAM, 0, td->td_ucred, td);
if (error != 0)
panic("%s: socreate, error=%d", __func__, error);
STAILQ_FOREACH(ifctx, &gctx->interfaces, next)
bootpc_fakeup_interface(ifctx, td);
STAILQ_FOREACH(ifctx, &gctx->interfaces, next)
bootpc_compose_query(ifctx, td);
error = bootpc_call(gctx, td);
2000-10-24 22:40:18 +00:00
if (error != 0) {
printf("BOOTP call failed\n");
}
mountopts(&nd->root_args, NULL);
STAILQ_FOREACH(ifctx, &gctx->interfaces, next)
if (bootpc_ifctx_isresolved(ifctx) != 0)
bootpc_decode_reply(nd, ifctx, gctx);
#ifdef BOOTP_NFSROOT
Changes to allow using BOOTP_NFSROOT and mounting an nfs root filesystem other than the one specified by the BOOTP server. This configures NFS using the BOOTP protocol while also respecting other root-path options such as setting vfs.root.mountfrom in the environment or using the RB_DFLTROOT boot option. It allows you to override the root path provided by the server, or to supply a root path when the server provides IP configuration but no root path info. This maintains the historical BOOTP_NFSROOT behavior of panicking on a failure to mount the root path provided by the server, unless you've provided an alternative via the ROOTDEVNAME kernel option or by setting vfs.root.mountfrom. The behavior of panicking when given no other options is preserved because it amounts to a bit of a retry loop that could eventually recover from a transient network or server problem. The user can now override the root path from loader(8) even if the kernel is compiled with BOOTP_NFSROOT. If vfs.root.mountfrom is set in the environment it is used unconditionally -- it always overrides the BOOTP info. If it begins with [old]nfs: then the BOOTP code uses it instead of the server-provided info. If it specifies some other filesystem then the bootp code will not panic like it used to and the code in vfs_mountroot.c will invoke the right filesystem to do the mount. If the kernel is compiled with the ROOTDEVNAME option, then that name is used by the BOOTP code if either * The server doesn't provide a pathname. * The boothowto flags include RB_DFLTROOT. The latter allows the user to compile in alternate path in ROOTDEVNAME such as ufs:/dev/da0s1a and boot from that path by setting boot_dftlroot=1 in loader(8) or using the '-r' option in boot(8). The one thing not provided here is automatic failover from a server-provided path to a compiled-in one without the user manually requesting that. The code just isn't currently structured in a way that makes that possible with a lot of rewrite. I think the ability to set vfs.root.mountfrom and to use ROOTDEVNAME automatically when the server doesn't provide a name covers the most common needs. A set of patches submitted by Lars Eggert provided the part I couldn't figure out by myself when I tried to do this last year; many thanks. Reviewed by: rodrigc
2013-07-31 19:14:00 +00:00
if (gctx->gotrootpath == 0 && gctx->any_root_overrides == 0)
2000-10-24 22:40:18 +00:00
panic("bootpc: No root path offered");
#endif
STAILQ_FOREACH(ifctx, &gctx->interfaces, next)
bootpc_adjust_interface(ifctx, gctx, td);
soclose(bootp_so);
STAILQ_FOREACH(ifctx, &gctx->interfaces, next)
if (ifctx->gotrootpath != 0)
break;
if (ifctx == NULL) {
STAILQ_FOREACH(ifctx, &gctx->interfaces, next)
if (bootpc_ifctx_isresolved(ifctx) != 0)
break;
}
if (ifctx == NULL)
goto out;
if (gctx->gotrootpath != 0) {
struct epoch_tracker et;
kern_setenv("boot.netif.name", ifctx->ifp->if_xname);
NET_EPOCH_ENTER(et);
bootpc_add_default_route(ifctx);
2000-10-24 22:40:18 +00:00
error = md_mount(&nd->root_saddr, nd->root_hostnam,
nd->root_fh, &nd->root_fhsize,
&nd->root_args, td);
bootpc_remove_default_route(ifctx);
NET_EPOCH_EXIT(et);
Changes to allow using BOOTP_NFSROOT and mounting an nfs root filesystem other than the one specified by the BOOTP server. This configures NFS using the BOOTP protocol while also respecting other root-path options such as setting vfs.root.mountfrom in the environment or using the RB_DFLTROOT boot option. It allows you to override the root path provided by the server, or to supply a root path when the server provides IP configuration but no root path info. This maintains the historical BOOTP_NFSROOT behavior of panicking on a failure to mount the root path provided by the server, unless you've provided an alternative via the ROOTDEVNAME kernel option or by setting vfs.root.mountfrom. The behavior of panicking when given no other options is preserved because it amounts to a bit of a retry loop that could eventually recover from a transient network or server problem. The user can now override the root path from loader(8) even if the kernel is compiled with BOOTP_NFSROOT. If vfs.root.mountfrom is set in the environment it is used unconditionally -- it always overrides the BOOTP info. If it begins with [old]nfs: then the BOOTP code uses it instead of the server-provided info. If it specifies some other filesystem then the bootp code will not panic like it used to and the code in vfs_mountroot.c will invoke the right filesystem to do the mount. If the kernel is compiled with the ROOTDEVNAME option, then that name is used by the BOOTP code if either * The server doesn't provide a pathname. * The boothowto flags include RB_DFLTROOT. The latter allows the user to compile in alternate path in ROOTDEVNAME such as ufs:/dev/da0s1a and boot from that path by setting boot_dftlroot=1 in loader(8) or using the '-r' option in boot(8). The one thing not provided here is automatic failover from a server-provided path to a compiled-in one without the user manually requesting that. The code just isn't currently structured in a way that makes that possible with a lot of rewrite. I think the ability to set vfs.root.mountfrom and to use ROOTDEVNAME automatically when the server doesn't provide a name covers the most common needs. A set of patches submitted by Lars Eggert provided the part I couldn't figure out by myself when I tried to do this last year; many thanks. Reviewed by: rodrigc
2013-07-31 19:14:00 +00:00
if (error != 0) {
if (gctx->any_root_overrides == 0)
panic("nfs_boot: mount root, error=%d", error);
else
goto out;
}
rootdevnames[0] = "nfs:";
2000-10-24 22:40:18 +00:00
nfs_diskless_valid = 3;
}
strcpy(nd->myif.ifra_name, ifctx->ireq.ifr_name);
bcopy(&ifctx->myaddr, &nd->myif.ifra_addr, sizeof(ifctx->myaddr));
bcopy(&ifctx->myaddr, &nd->myif.ifra_broadaddr, sizeof(ifctx->myaddr));
2000-10-24 22:40:18 +00:00
((struct sockaddr_in *) &nd->myif.ifra_broadaddr)->sin_addr.s_addr =
ifctx->myaddr.sin_addr.s_addr |
~ ifctx->netmask.sin_addr.s_addr;
bcopy(&ifctx->netmask, &nd->myif.ifra_mask, sizeof(ifctx->netmask));
bcopy(&ifctx->gw, &nd->mygateway, sizeof(ifctx->gw));
out:
while((ifctx = STAILQ_FIRST(&gctx->interfaces)) != NULL) {
STAILQ_REMOVE_HEAD(&gctx->interfaces, next);
free(ifctx, M_TEMP);
}
free(gctx, M_TEMP);
}
/*
* RPC: mountd/mount
* Given a server pathname, get an NFS file handle.
* Also, sets sin->sin_port to the NFS service port.
*/
static int
md_mount(struct sockaddr_in *mdsin, char *path, u_char *fhp, int *fhsizep,
struct nfs_args *args, struct thread *td)
{
struct mbuf *m;
int error;
int authunixok;
int authcount;
int authver;
#define RPCPROG_MNT 100005
#define RPCMNT_VER1 1
#define RPCMNT_VER3 3
#define RPCMNT_MOUNT 1
#define AUTH_SYS 1 /* unix style (uid, gids) */
#define AUTH_UNIX AUTH_SYS
/* XXX honor v2/v3 flags in args->flags? */
#ifdef BOOTP_NFSV3
/* First try NFS v3 */
/* Get port number for MOUNTD. */
error = krpc_portmap(mdsin, RPCPROG_MNT, RPCMNT_VER3,
&mdsin->sin_port, td);
if (error == 0) {
m = xdr_string_encode(path, strlen(path));
/* Do RPC to mountd. */
error = krpc_call(mdsin, RPCPROG_MNT, RPCMNT_VER3,
RPCMNT_MOUNT, &m, NULL, td);
}
if (error == 0) {
args->flags |= NFSMNT_NFSV3;
} else {
#endif
2000-10-24 22:40:18 +00:00
/* Fallback to NFS v2 */
2000-10-24 22:40:18 +00:00
/* Get port number for MOUNTD. */
error = krpc_portmap(mdsin, RPCPROG_MNT, RPCMNT_VER1,
&mdsin->sin_port, td);
2000-10-24 22:40:18 +00:00
if (error != 0)
return error;
2000-10-24 22:40:18 +00:00
m = xdr_string_encode(path, strlen(path));
2000-10-24 22:40:18 +00:00
/* Do RPC to mountd. */
error = krpc_call(mdsin, RPCPROG_MNT, RPCMNT_VER1,
RPCMNT_MOUNT, &m, NULL, td);
2000-10-24 22:40:18 +00:00
if (error != 0)
return error; /* message already freed */
#ifdef BOOTP_NFSV3
}
#endif
2000-10-24 22:40:18 +00:00
if (xdr_int_decode(&m, &error) != 0 || error != 0)
goto bad;
2000-10-24 22:40:18 +00:00
if ((args->flags & NFSMNT_NFSV3) != 0) {
if (xdr_int_decode(&m, fhsizep) != 0 ||
*fhsizep > NFSX_V3FHMAX ||
*fhsizep <= 0)
goto bad;
} else
*fhsizep = NFSX_V2FH;
2000-10-24 22:40:18 +00:00
if (xdr_opaque_decode(&m, fhp, *fhsizep) != 0)
goto bad;
if (args->flags & NFSMNT_NFSV3) {
2000-10-24 22:40:18 +00:00
if (xdr_int_decode(&m, &authcount) != 0)
goto bad;
authunixok = 0;
if (authcount < 0 || authcount > 100)
goto bad;
while (authcount > 0) {
if (xdr_int_decode(&m, &authver) != 0)
goto bad;
if (authver == AUTH_UNIX)
2000-10-24 22:40:18 +00:00
authunixok = 1;
authcount--;
}
if (authunixok == 0)
goto bad;
}
/* Set port number for NFS use. */
2000-10-24 22:40:18 +00:00
error = krpc_portmap(mdsin, NFS_PROG,
(args->flags &
NFSMNT_NFSV3) ? NFS_VER3 : NFS_VER2,
&mdsin->sin_port, td);
goto out;
bad:
error = EBADRPC;
out:
m_freem(m);
return error;
}
SYSINIT(bootp_rootconf, SI_SUB_ROOT_CONF, SI_ORDER_FIRST, bootpc_init, NULL);