linux/fs/nfs/nfsroot.c
Chuck Lever 53d4737580 NFS: NFSROOT should default to "proto=udp"
There have been a number of recent reports that NFSROOT is no longer
working with default mount options, but fails only with certain NICs.

Brian Downing <bdowning@lavos.net> bisected to commit 56463e50 "NFS:
Use super.c for NFSROOT mount option parsing".  Among other things,
this commit changes the default mount options for NFSROOT to use TCP
instead of UDP as the underlying transport.

TCP seems less able to deal with NICs that are slow to initialize.
The system logs that have accompanied reports of problems all show
that NFSROOT attempts to establish a TCP connection before the NIC is
fully initialized, and thus the TCP connection attempt fails.

When a TCP connection attempt fails during a mount operation, the
NFS stack needs to fail the operation.  Usually user space knows how
and when to retry it.  The network layer does not report a distinct
error code for this particular failure mode.  Thus, there isn't a
clean way for the RPC client to see that it needs to retry in this
case, but not in others.

Because NFSROOT is used in some environments where it is not possible
to update the kernel command line to specify "udp", the proper thing
to do is change NFSROOT to use UDP by default, as it did before commit
56463e50.

To make it easier to see how to change default mount options for
NFSROOT and to distinguish default settings from mandatory settings,
I've adjusted a couple of areas to document the specifics.

root_nfs_cat() is also modified to deal with commas properly when
concatenating strings containing mount option lists.  This keeps
root_nfs_cat() call sites simpler, now that we may be concatenating
multiple mount option strings.

Tested-by: Brian Downing <bdowning@lavos.net>
Tested-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Cc: <stable@kernel.org> # 2.6.37
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2011-03-11 15:38:07 -05:00

310 lines
9.4 KiB
C

/*
* Copyright (C) 1995, 1996 Gero Kuhlmann <gero@gkminix.han.de>
*
* Allow an NFS filesystem to be mounted as root. The way this works is:
* (1) Use the IP autoconfig mechanism to set local IP addresses and routes.
* (2) Construct the device string and the options string using DHCP
* option 17 and/or kernel command line options.
* (3) When mount_root() sets up the root file system, pass these strings
* to the NFS client's regular mount interface via sys_mount().
*
*
* Changes:
*
* Alan Cox : Removed get_address name clash with FPU.
* Alan Cox : Reformatted a bit.
* Gero Kuhlmann : Code cleanup
* Michael Rausch : Fixed recognition of an incoming RARP answer.
* Martin Mares : (2.0) Auto-configuration via BOOTP supported.
* Martin Mares : Manual selection of interface & BOOTP/RARP.
* Martin Mares : Using network routes instead of host routes,
* allowing the default configuration to be used
* for normal operation of the host.
* Martin Mares : Randomized timer with exponential backoff
* installed to minimize network congestion.
* Martin Mares : Code cleanup.
* Martin Mares : (2.1) BOOTP and RARP made configuration options.
* Martin Mares : Server hostname generation fixed.
* Gerd Knorr : Fixed wired inode handling
* Martin Mares : (2.2) "0.0.0.0" addresses from command line ignored.
* Martin Mares : RARP replies not tested for server address.
* Gero Kuhlmann : (2.3) Some bug fixes and code cleanup again (please
* send me your new patches _before_ bothering
* Linus so that I don' always have to cleanup
* _afterwards_ - thanks)
* Gero Kuhlmann : Last changes of Martin Mares undone.
* Gero Kuhlmann : RARP replies are tested for specified server
* again. However, it's now possible to have
* different RARP and NFS servers.
* Gero Kuhlmann : "0.0.0.0" addresses from command line are
* now mapped to INADDR_NONE.
* Gero Kuhlmann : Fixed a bug which prevented BOOTP path name
* from being used (thanks to Leo Spiekman)
* Andy Walker : Allow to specify the NFS server in nfs_root
* without giving a path name
* Swen Thümmler : Allow to specify the NFS options in nfs_root
* without giving a path name. Fix BOOTP request
* for domainname (domainname is NIS domain, not
* DNS domain!). Skip dummy devices for BOOTP.
* Jacek Zapala : Fixed a bug which prevented server-ip address
* from nfsroot parameter from being used.
* Olaf Kirch : Adapted to new NFS code.
* Jakub Jelinek : Free used code segment.
* Marko Kohtala : Fixed some bugs.
* Martin Mares : Debug message cleanup
* Martin Mares : Changed to use the new generic IP layer autoconfig
* code. BOOTP and RARP moved there.
* Martin Mares : Default path now contains host name instead of
* host IP address (but host name defaults to IP
* address anyway).
* Martin Mares : Use root_server_addr appropriately during setup.
* Martin Mares : Rewrote parameter parsing, now hopefully giving
* correct overriding.
* Trond Myklebust : Add in preliminary support for NFSv3 and TCP.
* Fix bug in root_nfs_addr(). nfs_data.namlen
* is NOT for the length of the hostname.
* Hua Qin : Support for mounting root file system via
* NFS over TCP.
* Fabian Frederick: Option parser rebuilt (using parser lib)
* Chuck Lever : Use super.c's text-based mount option parsing
* Chuck Lever : Add "nfsrootdebug".
*/
#include <linux/types.h>
#include <linux/string.h>
#include <linux/init.h>
#include <linux/nfs.h>
#include <linux/nfs_fs.h>
#include <linux/utsname.h>
#include <linux/root_dev.h>
#include <net/ipconfig.h>
#include "internal.h"
#define NFSDBG_FACILITY NFSDBG_ROOT
/* Default path we try to mount. "%s" gets replaced by our IP address */
#define NFS_ROOT "/tftpboot/%s"
/* Default NFSROOT mount options. */
#define NFS_DEF_OPTIONS "udp"
/* Parameters passed from the kernel command line */
static char nfs_root_parms[256] __initdata = "";
/* Text-based mount options passed to super.c */
static char nfs_root_options[256] __initdata = NFS_DEF_OPTIONS;
/* Address of NFS server */
static __be32 servaddr __initdata = htonl(INADDR_NONE);
/* Name of directory to mount */
static char nfs_export_path[NFS_MAXPATHLEN + 1] __initdata = "";
/* server:export path string passed to super.c */
static char nfs_root_device[NFS_MAXPATHLEN + 1] __initdata = "";
#ifdef RPC_DEBUG
/*
* When the "nfsrootdebug" kernel command line option is specified,
* enable debugging messages for NFSROOT.
*/
static int __init nfs_root_debug(char *__unused)
{
nfs_debug |= NFSDBG_ROOT | NFSDBG_MOUNT;
return 1;
}
__setup("nfsrootdebug", nfs_root_debug);
#endif
/*
* Parse NFS server and directory information passed on the kernel
* command line.
*
* nfsroot=[<server-ip>:]<root-dir>[,<nfs-options>]
*
* If there is a "%s" token in the <root-dir> string, it is replaced
* by the ASCII-representation of the client's IP address.
*/
static int __init nfs_root_setup(char *line)
{
ROOT_DEV = Root_NFS;
if (line[0] == '/' || line[0] == ',' || (line[0] >= '0' && line[0] <= '9')) {
strlcpy(nfs_root_parms, line, sizeof(nfs_root_parms));
} else {
size_t n = strlen(line) + sizeof(NFS_ROOT) - 1;
if (n >= sizeof(nfs_root_parms))
line[sizeof(nfs_root_parms) - sizeof(NFS_ROOT) - 2] = '\0';
sprintf(nfs_root_parms, NFS_ROOT, line);
}
/*
* Extract the IP address of the NFS server containing our
* root file system, if one was specified.
*
* Note: root_nfs_parse_addr() removes the server-ip from
* nfs_root_parms, if it exists.
*/
root_server_addr = root_nfs_parse_addr(nfs_root_parms);
return 1;
}
__setup("nfsroot=", nfs_root_setup);
static int __init root_nfs_copy(char *dest, const char *src,
const size_t destlen)
{
if (strlcpy(dest, src, destlen) > destlen)
return -1;
return 0;
}
static int __init root_nfs_cat(char *dest, const char *src,
const size_t destlen)
{
size_t len = strlen(dest);
if (len && dest[len - 1] != ',')
if (strlcat(dest, ",", destlen) > destlen)
return -1;
if (strlcat(dest, src, destlen) > destlen)
return -1;
return 0;
}
/*
* Parse out root export path and mount options from
* passed-in string @incoming.
*
* Copy the export path into @exppath.
*/
static int __init root_nfs_parse_options(char *incoming, char *exppath,
const size_t exppathlen)
{
char *p;
/*
* Set the NFS remote path
*/
p = strsep(&incoming, ",");
if (*p != '\0' && strcmp(p, "default") != 0)
if (root_nfs_copy(exppath, p, exppathlen))
return -1;
/*
* @incoming now points to the rest of the string; if it
* contains something, append it to our root options buffer
*/
if (incoming != NULL && *incoming != '\0')
if (root_nfs_cat(nfs_root_options, incoming,
sizeof(nfs_root_options)))
return -1;
return 0;
}
/*
* Decode the export directory path name and NFS options from
* the kernel command line. This has to be done late in order to
* use a dynamically acquired client IP address for the remote
* root directory path.
*
* Returns zero if successful; otherwise -1 is returned.
*/
static int __init root_nfs_data(char *cmdline)
{
char mand_options[sizeof("nolock,addr=") + INET_ADDRSTRLEN + 1];
int len, retval = -1;
char *tmp = NULL;
const size_t tmplen = sizeof(nfs_export_path);
tmp = kzalloc(tmplen, GFP_KERNEL);
if (tmp == NULL)
goto out_nomem;
strcpy(tmp, NFS_ROOT);
if (root_server_path[0] != '\0') {
dprintk("Root-NFS: DHCPv4 option 17: %s\n",
root_server_path);
if (root_nfs_parse_options(root_server_path, tmp, tmplen))
goto out_optionstoolong;
}
if (cmdline[0] != '\0') {
dprintk("Root-NFS: nfsroot=%s\n", cmdline);
if (root_nfs_parse_options(cmdline, tmp, tmplen))
goto out_optionstoolong;
}
/*
* Append mandatory options for nfsroot so they override
* what has come before
*/
snprintf(mand_options, sizeof(mand_options), "nolock,addr=%pI4",
&servaddr);
if (root_nfs_cat(nfs_root_options, mand_options,
sizeof(nfs_root_options)))
goto out_optionstoolong;
/*
* Set up nfs_root_device. For NFS mounts, this looks like
*
* server:/path
*
* At this point, utsname()->nodename contains our local
* IP address or hostname, set by ipconfig. If "%s" exists
* in tmp, substitute the nodename, then shovel the whole
* mess into nfs_root_device.
*/
len = snprintf(nfs_export_path, sizeof(nfs_export_path),
tmp, utsname()->nodename);
if (len > (int)sizeof(nfs_export_path))
goto out_devnametoolong;
len = snprintf(nfs_root_device, sizeof(nfs_root_device),
"%pI4:%s", &servaddr, nfs_export_path);
if (len > (int)sizeof(nfs_root_device))
goto out_devnametoolong;
retval = 0;
out:
kfree(tmp);
return retval;
out_nomem:
printk(KERN_ERR "Root-NFS: could not allocate memory\n");
goto out;
out_optionstoolong:
printk(KERN_ERR "Root-NFS: mount options string too long\n");
goto out;
out_devnametoolong:
printk(KERN_ERR "Root-NFS: root device name too long.\n");
goto out;
}
/**
* nfs_root_data - Return prepared 'data' for NFSROOT mount
* @root_device: OUT: address of string containing NFSROOT device
* @root_data: OUT: address of string containing NFSROOT mount options
*
* Returns zero and sets @root_device and @root_data if successful,
* otherwise -1 is returned.
*/
int __init nfs_root_data(char **root_device, char **root_data)
{
servaddr = root_server_addr;
if (servaddr == htonl(INADDR_NONE)) {
printk(KERN_ERR "Root-NFS: no NFS server address\n");
return -1;
}
if (root_nfs_data(nfs_root_parms) < 0)
return -1;
*root_device = nfs_root_device;
*root_data = nfs_root_options;
return 0;
}