[Virtio-fs] [RFC PATCH] virtiofsd: Provide support for posix locks
Liu Bo
bo.liu at linux.alibaba.com
Thu May 30 19:00:21 UTC 2019
On Thu, May 30, 2019 at 02:50:05PM -0400, Vivek Goyal wrote:
> On Thu, May 30, 2019 at 11:45:11AM -0700, Liu Bo wrote:
> > On Thu, May 30, 2019 at 02:11:41PM -0400, Vivek Goyal wrote:
> > > Doing posix locks with-in guest kernel are not sufficient if a file/dir
> > > is being shared by multiple guests. So we need the notion of daemon doing
> > > the locks which are visible to rest of the guests.
> > >
> > > Given posix locks are per process, one can not call posix lock API on host,
> > > otherwise bunch of basic posix locks properties are broken. For example,
> > > If two processes (A and B) in guest open the file and take locks on different
> > > sections of file, if one of the processes closes the fd, it will close
> > > fd on virtiofsd and all posix locks on file will go away. This means if
> > > process A closes the fd, then locks of process B will go away too.
> > >
> > > Similar other problems exist too.
> > >
> > > This patch set tries to emulate posix locks while using open file
> > > description locks provided on Linux.
> > >
> > > Daemon provides two options (-o posix_lock, -o no_posix_lock) to enable
> > > or disable posix locking in daemon. By default it is enabled.
> > >
> > > There are few issues though.
> > >
> > > - GETLK() returns pid of process holding lock. As we are emulating locks
> > > using OFD, and these locks are not per process and don't return pid
> > > of process, so GETLK() in guest does not reuturn process pid.
> > >
> > > - As of now only F_SETLK is supported and not F_SETLKW. We can't block
> > > the thread in virtiofsd for arbitrary long duration as there is only
> > > one thread serving the queue. That means unlock request will not make
> > > it to daemon and F_SETLKW will block infinitely and bring virtio-fs
> > > to a halt. This is a solvable problem though and will require significant
> > > changes in virtiofsd and kernel. Left as a TODO item for now.
> >
> > We've also seen this hang with flock()'s sleep mode, I was wondering
> > if we could pthread_create a new thread to do the sleeping locking.
>
> One idea I was discussing with david gilbert is, can we have multiple
> threds serving same virt queue and that will allow us blocking in same
> context as the caller.
>
> This probably also means create a separate virtqueue for sending down
> requests which can block for arbitrarily long amount of time, to make
> sure deadlock does not happen.
>
Right, with "separate virtqueue" + "multi threading" together, the
problem should be addressed.
thanks,
-liubo
> Thanks
> Vivek
>
> >
> > thanks,
> > -liubo
> > >
> > > Signed-off-by: Vivek Goyal <vgoyal at redhat.com>
> > > ---
> > > contrib/virtiofsd/passthrough_ll.c | 185 ++++++++++++++++++++++++++++++++++++-
> > > 1 file changed, 184 insertions(+), 1 deletion(-)
> > >
> > > Index: qemu/contrib/virtiofsd/passthrough_ll.c
> > > ===================================================================
> > > --- qemu.orig/contrib/virtiofsd/passthrough_ll.c 2019-04-25 10:49:14.103386416 -0400
> > > +++ qemu/contrib/virtiofsd/passthrough_ll.c 2019-05-30 14:02:55.598483536 -0400
> > > @@ -58,6 +58,12 @@
> > > #include <gmodule.h>
> > > #include "seccomp.h"
> > >
> > > +/* Keep track of inode posix locks for each owner. */
> > > +struct lo_inode_plock {
> > > + uint64_t lock_owner;
> > > + int fd; /* fd for OFD locks */
> > > +};
> > > +
> > > struct lo_map_elem {
> > > union {
> > > struct lo_inode *inode;
> > > @@ -86,6 +92,8 @@ struct lo_inode {
> > > struct lo_key key;
> > > uint64_t refcount; /* protected by lo->mutex */
> > > fuse_ino_t fuse_ino;
> > > + pthread_mutex_t mutex;
> > > + GHashTable *posix_locks; /* protected by lo_inode->mutex */
> > > };
> > >
> > > struct lo_cred {
> > > @@ -105,6 +113,7 @@ struct lo_data {
> > > int norace;
> > > int writeback;
> > > int flock;
> > > + int posix_lock;
> > > int xattr;
> > > const char *source;
> > > double timeout;
> > > @@ -133,6 +142,10 @@ static const struct fuse_opt lo_opts[] =
> > > offsetof(struct lo_data, flock), 1 },
> > > { "no_flock",
> > > offsetof(struct lo_data, flock), 0 },
> > > + { "posix_lock",
> > > + offsetof(struct lo_data, posix_lock), 0 },
> > > + { "no_posix_lock",
> > > + offsetof(struct lo_data, posix_lock), 0 },
> > > { "xattr",
> > > offsetof(struct lo_data, xattr), 1 },
> > > { "no_xattr",
> > > @@ -362,13 +375,24 @@ static void lo_init(void *userdata,
> > > fprintf(stderr, "lo_init: activating flock locks\n");
> > > conn->want |= FUSE_CAP_FLOCK_LOCKS;
> > > }
> > > +
> > > + if (conn->capable & FUSE_CAP_POSIX_LOCKS) {
> > > + if (lo->posix_lock) {
> > > + if (lo->debug)
> > > + fprintf(stderr, "lo_init: activating posix locks\n");
> > > + conn->want |= FUSE_CAP_POSIX_LOCKS;
> > > + } else {
> > > + if (lo->debug)
> > > + fprintf(stderr, "lo_init: disabling posix locks\n");
> > > + conn->want &= ~FUSE_CAP_POSIX_LOCKS;
> > > + }
> > > + }
> > > if ((lo->cache == CACHE_NONE && !lo->readdirplus_set) ||
> > > lo->readdirplus_clear) {
> > > if (lo->debug)
> > > fprintf(stderr, "lo_init: disabling readdirplus\n");
> > > conn->want &= ~FUSE_CAP_READDIRPLUS;
> > > }
> > > -
> > > }
> > >
> > > static void lo_getattr(fuse_req_t req, fuse_ino_t ino,
> > > @@ -673,6 +697,8 @@ static int lo_do_lookup(fuse_req_t req,
> > > newfd = -1;
> > > inode->key.ino = e->attr.st_ino;
> > > inode->key.dev = e->attr.st_dev;
> > > + pthread_mutex_init(&inode->mutex, NULL);
> > > + inode->posix_locks = g_hash_table_new(g_direct_hash, g_direct_equal);
> > >
> > > pthread_mutex_lock(&lo->mutex);
> > > inode->fuse_ino = lo_add_inode_mapping(req, inode);
> > > @@ -1038,6 +1064,10 @@ static void unref_inode(struct lo_data *
> > > if (!inode->refcount) {
> > > lo_map_remove(&lo->ino_map, inode->fuse_ino);
> > > g_hash_table_remove(lo->inodes, &inode->key);
> > > + if (g_hash_table_size(inode->posix_locks)) {
> > > + warn("Hash table is not empty\n");
> > > + }
> > > + g_hash_table_destroy(inode->posix_locks);
> > > pthread_mutex_unlock(&lo->mutex);
> > > close(inode->fd);
> > > free(inode);
> > > @@ -1379,6 +1409,131 @@ out:
> > > fuse_reply_create(req, &e, fi);
> > > }
> > >
> > > +/* Should be called with inode->mutex held */
> > > +static struct lo_inode_plock *lookup_create_plock_ctx(struct lo_data *lo,
> > > + struct lo_inode *inode, uint64_t lock_owner,
> > > + pid_t pid, int *err)
> > > +{
> > > + struct lo_inode_plock *plock;
> > > + char procname[64];
> > > + int fd;
> > > +
> > > + plock = g_hash_table_lookup(inode->posix_locks,
> > > + GUINT_TO_POINTER(lock_owner));
> > > +
> > > + if (plock)
> > > + return plock;
> > > +
> > > + plock = malloc(sizeof(struct lo_inode_plock));
> > > + if (!plock) {
> > > + *err = ENOMEM;
> > > + return NULL;
> > > + }
> > > +
> > > + /* Open another instance of file which can be used for ofd locks. */
> > > + sprintf(procname, "%i", inode->fd);
> > > +
> > > + /* TODO: What if file is not writable? */
> > > + fd = openat(lo->proc_self_fd, procname, O_RDWR);
> > > + if (fd == -1) {
> > > + *err = -errno;
> > > + free(plock);
> > > + return NULL;
> > > + }
> > > +
> > > + plock->lock_owner = lock_owner;
> > > + plock->fd = fd;
> > > + g_hash_table_insert(inode->posix_locks,
> > > + GUINT_TO_POINTER(plock->lock_owner), plock);
> > > + return plock;
> > > +}
> > > +
> > > +static void lo_getlk(fuse_req_t req, fuse_ino_t ino,
> > > + struct fuse_file_info *fi, struct flock *lock)
> > > +{
> > > + struct lo_data *lo = lo_data(req);
> > > + struct lo_inode *inode;
> > > + struct lo_inode_plock *plock;
> > > + int ret, saverr = 0;
> > > +
> > > + if (lo_debug(req))
> > > + fprintf(stderr, "lo_getlk(ino=%" PRIu64 ", flags=%d)"
> > > + " owner=0x%lx, l_type=%d l_start=0x%lx"
> > > + " l_len=0x%lx\n", ino, fi->flags, fi->lock_owner,
> > > + lock->l_type, lock->l_start, lock->l_len);
> > > +
> > > + inode = lo_inode(req, ino);
> > > + if (!inode) {
> > > + fuse_reply_err(req, EBADF);
> > > + return;
> > > + }
> > > +
> > > + pthread_mutex_lock(&inode->mutex);
> > > + plock = lookup_create_plock_ctx(lo, inode, fi->lock_owner, lock->l_pid,
> > > + &ret);
> > > + if (!plock) {
> > > + pthread_mutex_unlock(&inode->mutex);
> > > + fuse_reply_err(req, ret);
> > > + return;
> > > + }
> > > +
> > > + ret = fcntl(plock->fd, F_OFD_GETLK, lock);
> > > + if (ret == -1)
> > > + saverr = errno;
> > > + pthread_mutex_unlock(&inode->mutex);
> > > +
> > > + if (saverr)
> > > + fuse_reply_err(req, saverr);
> > > + else
> > > + fuse_reply_lock(req, lock);
> > > +}
> > > +
> > > +static void lo_setlk(fuse_req_t req, fuse_ino_t ino,
> > > + struct fuse_file_info *fi, struct flock *lock, int sleep)
> > > +{
> > > + struct lo_data *lo = lo_data(req);
> > > + struct lo_inode *inode;
> > > + struct lo_inode_plock *plock;
> > > + int ret, saverr = 0;
> > > +
> > > + if (lo_debug(req))
> > > + fprintf(stderr, "lo_setlk(ino=%" PRIu64 ", flags=%d)"
> > > + " cmd=%d pid=%d owner=0x%lx sleep=%d l_whence=%d"
> > > + " l_start=0x%lx l_len=0x%lx\n", ino, fi->flags,
> > > + lock->l_type, lock->l_pid, fi->lock_owner, sleep,
> > > + lock->l_whence, lock->l_start, lock->l_len);
> > > +
> > > + if (sleep) {
> > > + fuse_reply_err(req, EOPNOTSUPP);
> > > + return;
> > > + }
> > > +
> > > + inode = lo_inode(req, ino);
> > > + if (!inode) {
> > > + fuse_reply_err(req, EBADF);
> > > + return;
> > > + }
> > > +
> > > + pthread_mutex_lock(&inode->mutex);
> > > + plock = lookup_create_plock_ctx(lo, inode, fi->lock_owner, lock->l_pid,
> > > + &ret);
> > > +
> > > + if (!plock) {
> > > + pthread_mutex_unlock(&inode->mutex);
> > > + fuse_reply_err(req, ret);
> > > + return;
> > > + }
> > > +
> > > + /* TODO: Is it alright to modify flock? */
> > > + lock->l_pid = 0;
> > > + ret = fcntl(plock->fd, F_OFD_SETLK, lock);
> > > + if (ret == -1) {
> > > + saverr = errno;
> > > + }
> > > + pthread_mutex_unlock(&inode->mutex);
> > > + fuse_reply_err(req, saverr);
> > > +}
> > > +
> > > static void lo_fsyncdir(fuse_req_t req, fuse_ino_t ino, int datasync,
> > > struct fuse_file_info *fi)
> > > {
> > > @@ -1476,6 +1631,31 @@ static void lo_flush(fuse_req_t req, fus
> > > {
> > > int res;
> > > (void) ino;
> > > + struct lo_inode *inode;
> > > + struct lo_inode_plock *plock;
> > > +
> > > + inode = lo_inode(req, ino);
> > > + if (!inode) {
> > > + fuse_reply_err(req, EBADF);
> > > + return;
> > > + }
> > > +
> > > + /* An fd is going away. Cleanup associated posix locks */
> > > + pthread_mutex_lock(&inode->mutex);
> > > + plock = g_hash_table_lookup(inode->posix_locks,
> > > + GUINT_TO_POINTER(fi->lock_owner));
> > > + if (plock) {
> > > + g_hash_table_remove(inode->posix_locks,
> > > + GUINT_TO_POINTER(fi->lock_owner));
> > > + /*
> > > + * We had used open() for locks and had only one fd. So
> > > + * closing this fd should release all OFD locks.
> > > + */
> > > + close(plock->fd);
> > > + free(plock);
> > > + }
> > > + pthread_mutex_unlock(&inode->mutex);
> > > +
> > > res = close(dup(lo_fi_fd(req, fi)));
> > > fuse_reply_err(req, res == -1 ? errno : 0);
> > > }
> > > @@ -1963,6 +2143,8 @@ static struct fuse_lowlevel_ops lo_oper
> > > .releasedir = lo_releasedir,
> > > .fsyncdir = lo_fsyncdir,
> > > .create = lo_create,
> > > + .getlk = lo_getlk,
> > > + .setlk = lo_setlk,
> > > .open = lo_open,
> > > .release = lo_release,
> > > .flush = lo_flush,
> > > @@ -2189,6 +2371,7 @@ int main(int argc, char *argv[])
> > > struct fuse_cmdline_opts opts;
> > > struct lo_data lo = { .debug = 0,
> > > .writeback = 0,
> > > + .posix_lock = 1,
> > > .proc_self_fd = -1,
> > > };
> > > struct lo_map_elem *root_elem;
> > >
> > > _______________________________________________
> > > Virtio-fs mailing list
> > > Virtio-fs at redhat.com
> > > https://www.redhat.com/mailman/listinfo/virtio-fs
More information about the Virtio-fs
mailing list