alistair23-linux/fs/lockd/clntlock.c
Linus Torvalds 700a800a94 This pull consists mostly of nfsd container work:
Scott Mayhew revived an old api that communicates with a userspace
 daemon to manage some on-disk state that's used to track clients across
 server reboots.  We've been using a usermode_helper upcall for that, but
 it's tough to run those with the right namespaces, so a daemon is much
 friendlier to container use cases.
 
 Trond fixed nfsd's handling of user credentials in user namespaces.  He
 also contributed patches that allow containers to support different sets
 of NFS protocol versions.
 
 The only remaining container bug I'm aware of is that the NFS reply
 cache is shared between all containers.  If anyone's aware of other gaps
 in our container support, let me know.
 
 The rest of this is miscellaneous bugfixes.
 -----BEGIN PGP SIGNATURE-----
 
 iQJJBAABCAAzFiEEYtFWavXG9hZotryuJ5vNeUKO4b4FAlzcWNcVHGJmaWVsZHNA
 ZmllbGRzZXMub3JnAAoJECebzXlCjuG+DUEP/0WD3jKNAHFV3M5YQPAI9fz/iCND
 Db/A4oWP5qa6JmwmHe61il29QeGqkeFr/NPexgzM3Xw2E39d7RBXBeWyVDuqb0wr
 6SCXjXibTsuAHg11nR8Xf0P5Vej3rfGbG6up5lLCIDTEZxVpWoaBJnM8+3bewuCj
 XbeiDW54oiMbmDjon3MXqVAIF/z7LjorecJ+Yw5+0Jy7KZ6num9Kt8+fi7qkEfFd
 i5Bp9KWgzlTbJUJV4EX3ZKN3zlGkfOvjoo2kP3PODPVMB34W8jSLKkRSA1tDWYZg
 43WhBt5OODDlV6zpxSJXehYKIB4Ae469+RRaIL4F+ORRK+AzR0C/GTuOwJiG+P3J
 n95DX5WzX74nPOGQJgAvq4JNpZci85jM3jEK1TR2M7KiBDG5Zg+FTsPYVxx5Sgah
 Akl/pjLtHQPSdBbFGHn5TsXU+gqWNiKsKa9663tjxLb8ldmJun6JoQGkAEF9UJUn
 dzv0UxyHeHAblhSynY+WsUR+Xep9JDo/p5LyFK4if9Sd62KeA1uF/MFhAqpKZF81
 mrgRCqW4sD8aVTBNZI06pZzmcZx4TRr2o+Oj5KAXf6Yk6TJRSGfnQscoMMBsTLkw
 VK1rBQ/71TpjLHGZZZEx1YJrkVZAMmw2ty4DtK2f9jeKO13bWmUpc6UATzVufHKA
 C1rUZXJ5YioDbYDy
 =TUdw
 -----END PGP SIGNATURE-----

Merge tag 'nfsd-5.2' of git://linux-nfs.org/~bfields/linux

Pull nfsd updates from Bruce Fields:
 "This consists mostly of nfsd container work:

  Scott Mayhew revived an old api that communicates with a userspace
  daemon to manage some on-disk state that's used to track clients
  across server reboots. We've been using a usermode_helper upcall for
  that, but it's tough to run those with the right namespaces, so a
  daemon is much friendlier to container use cases.

  Trond fixed nfsd's handling of user credentials in user namespaces. He
  also contributed patches that allow containers to support different
  sets of NFS protocol versions.

  The only remaining container bug I'm aware of is that the NFS reply
  cache is shared between all containers. If anyone's aware of other
  gaps in our container support, let me know.

  The rest of this is miscellaneous bugfixes"

* tag 'nfsd-5.2' of git://linux-nfs.org/~bfields/linux: (23 commits)
  nfsd: update callback done processing
  locks: move checks from locks_free_lock() to locks_release_private()
  nfsd: fh_drop_write in nfsd_unlink
  nfsd: allow fh_want_write to be called twice
  nfsd: knfsd must use the container user namespace
  SUNRPC: rsi_parse() should use the current user namespace
  SUNRPC: Fix the server AUTH_UNIX userspace mappings
  lockd: Pass the user cred from knfsd when starting the lockd server
  SUNRPC: Temporary sockets should inherit the cred from their parent
  SUNRPC: Cache the process user cred in the RPC server listener
  nfsd: Allow containers to set supported nfs versions
  nfsd: Add custom rpcbind callbacks for knfsd
  SUNRPC: Allow further customisation of RPC program registration
  SUNRPC: Clean up generic dispatcher code
  SUNRPC: Add a callback to initialise server requests
  SUNRPC/nfs: Fix return value for nfs4_callback_compound()
  nfsd: handle legacy client tracking records sent by nfsdcld
  nfsd: re-order client tracking method selection
  nfsd: keep a tally of RECLAIM_COMPLETE operations when using nfsdcld
  nfsd: un-deprecate nfsdcld
  ...
2019-05-15 18:21:43 -07:00

299 lines
7.5 KiB
C

/*
* linux/fs/lockd/clntlock.c
*
* Lock handling for the client side NLM implementation
*
* Copyright (C) 1996, Olaf Kirch <okir@monad.swb.de>
*/
#include <linux/module.h>
#include <linux/types.h>
#include <linux/slab.h>
#include <linux/time.h>
#include <linux/nfs_fs.h>
#include <linux/sunrpc/addr.h>
#include <linux/sunrpc/svc.h>
#include <linux/lockd/lockd.h>
#include <linux/kthread.h>
#define NLMDBG_FACILITY NLMDBG_CLIENT
/*
* Local function prototypes
*/
static int reclaimer(void *ptr);
/*
* The following functions handle blocking and granting from the
* client perspective.
*/
/*
* This is the representation of a blocked client lock.
*/
struct nlm_wait {
struct list_head b_list; /* linked list */
wait_queue_head_t b_wait; /* where to wait on */
struct nlm_host * b_host;
struct file_lock * b_lock; /* local file lock */
unsigned short b_reclaim; /* got to reclaim lock */
__be32 b_status; /* grant callback status */
};
static LIST_HEAD(nlm_blocked);
static DEFINE_SPINLOCK(nlm_blocked_lock);
/**
* nlmclnt_init - Set up per-NFS mount point lockd data structures
* @nlm_init: pointer to arguments structure
*
* Returns pointer to an appropriate nlm_host struct,
* or an ERR_PTR value.
*/
struct nlm_host *nlmclnt_init(const struct nlmclnt_initdata *nlm_init)
{
struct nlm_host *host;
u32 nlm_version = (nlm_init->nfs_version == 2) ? 1 : 4;
int status;
status = lockd_up(nlm_init->net, nlm_init->cred);
if (status < 0)
return ERR_PTR(status);
host = nlmclnt_lookup_host(nlm_init->address, nlm_init->addrlen,
nlm_init->protocol, nlm_version,
nlm_init->hostname, nlm_init->noresvport,
nlm_init->net, nlm_init->cred);
if (host == NULL)
goto out_nohost;
if (host->h_rpcclnt == NULL && nlm_bind_host(host) == NULL)
goto out_nobind;
host->h_nlmclnt_ops = nlm_init->nlmclnt_ops;
return host;
out_nobind:
nlmclnt_release_host(host);
out_nohost:
lockd_down(nlm_init->net);
return ERR_PTR(-ENOLCK);
}
EXPORT_SYMBOL_GPL(nlmclnt_init);
/**
* nlmclnt_done - Release resources allocated by nlmclnt_init()
* @host: nlm_host structure reserved by nlmclnt_init()
*
*/
void nlmclnt_done(struct nlm_host *host)
{
struct net *net = host->net;
nlmclnt_release_host(host);
lockd_down(net);
}
EXPORT_SYMBOL_GPL(nlmclnt_done);
/*
* Queue up a lock for blocking so that the GRANTED request can see it
*/
struct nlm_wait *nlmclnt_prepare_block(struct nlm_host *host, struct file_lock *fl)
{
struct nlm_wait *block;
block = kmalloc(sizeof(*block), GFP_KERNEL);
if (block != NULL) {
block->b_host = host;
block->b_lock = fl;
init_waitqueue_head(&block->b_wait);
block->b_status = nlm_lck_blocked;
spin_lock(&nlm_blocked_lock);
list_add(&block->b_list, &nlm_blocked);
spin_unlock(&nlm_blocked_lock);
}
return block;
}
void nlmclnt_finish_block(struct nlm_wait *block)
{
if (block == NULL)
return;
spin_lock(&nlm_blocked_lock);
list_del(&block->b_list);
spin_unlock(&nlm_blocked_lock);
kfree(block);
}
/*
* Block on a lock
*/
int nlmclnt_block(struct nlm_wait *block, struct nlm_rqst *req, long timeout)
{
long ret;
/* A borken server might ask us to block even if we didn't
* request it. Just say no!
*/
if (block == NULL)
return -EAGAIN;
/* Go to sleep waiting for GRANT callback. Some servers seem
* to lose callbacks, however, so we're going to poll from
* time to time just to make sure.
*
* For now, the retry frequency is pretty high; normally
* a 1 minute timeout would do. See the comment before
* nlmclnt_lock for an explanation.
*/
ret = wait_event_interruptible_timeout(block->b_wait,
block->b_status != nlm_lck_blocked,
timeout);
if (ret < 0)
return -ERESTARTSYS;
/* Reset the lock status after a server reboot so we resend */
if (block->b_status == nlm_lck_denied_grace_period)
block->b_status = nlm_lck_blocked;
req->a_res.status = block->b_status;
return 0;
}
/*
* The server lockd has called us back to tell us the lock was granted
*/
__be32 nlmclnt_grant(const struct sockaddr *addr, const struct nlm_lock *lock)
{
const struct file_lock *fl = &lock->fl;
const struct nfs_fh *fh = &lock->fh;
struct nlm_wait *block;
__be32 res = nlm_lck_denied;
/*
* Look up blocked request based on arguments.
* Warning: must not use cookie to match it!
*/
spin_lock(&nlm_blocked_lock);
list_for_each_entry(block, &nlm_blocked, b_list) {
struct file_lock *fl_blocked = block->b_lock;
if (fl_blocked->fl_start != fl->fl_start)
continue;
if (fl_blocked->fl_end != fl->fl_end)
continue;
/*
* Careful! The NLM server will return the 32-bit "pid" that
* we put on the wire: in this case the lockowner "pid".
*/
if (fl_blocked->fl_u.nfs_fl.owner->pid != lock->svid)
continue;
if (!rpc_cmp_addr(nlm_addr(block->b_host), addr))
continue;
if (nfs_compare_fh(NFS_FH(locks_inode(fl_blocked->fl_file)), fh) != 0)
continue;
/* Alright, we found a lock. Set the return status
* and wake up the caller
*/
block->b_status = nlm_granted;
wake_up(&block->b_wait);
res = nlm_granted;
}
spin_unlock(&nlm_blocked_lock);
return res;
}
/*
* The following procedures deal with the recovery of locks after a
* server crash.
*/
/*
* Reclaim all locks on server host. We do this by spawning a separate
* reclaimer thread.
*/
void
nlmclnt_recovery(struct nlm_host *host)
{
struct task_struct *task;
if (!host->h_reclaiming++) {
nlm_get_host(host);
task = kthread_run(reclaimer, host, "%s-reclaim", host->h_name);
if (IS_ERR(task))
printk(KERN_ERR "lockd: unable to spawn reclaimer "
"thread. Locks for %s won't be reclaimed! "
"(%ld)\n", host->h_name, PTR_ERR(task));
}
}
static int
reclaimer(void *ptr)
{
struct nlm_host *host = (struct nlm_host *) ptr;
struct nlm_wait *block;
struct nlm_rqst *req;
struct file_lock *fl, *next;
u32 nsmstate;
struct net *net = host->net;
req = kmalloc(sizeof(*req), GFP_KERNEL);
if (!req)
return 0;
allow_signal(SIGKILL);
down_write(&host->h_rwsem);
lockd_up(net, NULL); /* note: this cannot fail as lockd is already running */
dprintk("lockd: reclaiming locks for host %s\n", host->h_name);
restart:
nsmstate = host->h_nsmstate;
/* Force a portmap getport - the peer's lockd will
* most likely end up on a different port.
*/
host->h_nextrebind = jiffies;
nlm_rebind_host(host);
/* First, reclaim all locks that have been granted. */
list_splice_init(&host->h_granted, &host->h_reclaim);
list_for_each_entry_safe(fl, next, &host->h_reclaim, fl_u.nfs_fl.list) {
list_del_init(&fl->fl_u.nfs_fl.list);
/*
* sending this thread a SIGKILL will result in any unreclaimed
* locks being removed from the h_granted list. This means that
* the kernel will not attempt to reclaim them again if a new
* reclaimer thread is spawned for this host.
*/
if (signalled())
continue;
if (nlmclnt_reclaim(host, fl, req) != 0)
continue;
list_add_tail(&fl->fl_u.nfs_fl.list, &host->h_granted);
if (host->h_nsmstate != nsmstate) {
/* Argh! The server rebooted again! */
goto restart;
}
}
host->h_reclaiming = 0;
up_write(&host->h_rwsem);
dprintk("NLM: done reclaiming locks for host %s\n", host->h_name);
/* Now, wake up all processes that sleep on a blocked lock */
spin_lock(&nlm_blocked_lock);
list_for_each_entry(block, &nlm_blocked, b_list) {
if (block->b_host == host) {
block->b_status = nlm_lck_denied_grace_period;
wake_up(&block->b_wait);
}
}
spin_unlock(&nlm_blocked_lock);
/* Release host handle after use */
nlmclnt_release_host(host);
lockd_down(net);
kfree(req);
return 0;
}