Commit 5c672ab3 authored by Miklos Szeredi's avatar Miklos Szeredi

fuse: serialize dirops by default

Negotiate with userspace filesystems whether they support parallel readdir
and lookup.  Disable parallelism by default for fear of breaking fuse
filesystems.
Signed-off-by: default avatarMiklos Szeredi <mszeredi@redhat.com>
Fixes: 9902af79 ("parallel lookups: actual switch to rwsem")
Fixes: d9b3dbdc ("fuse: switch to ->iterate_shared()")
parent 4c2e07c6
...@@ -341,8 +341,10 @@ static struct dentry *fuse_lookup(struct inode *dir, struct dentry *entry, ...@@ -341,8 +341,10 @@ static struct dentry *fuse_lookup(struct inode *dir, struct dentry *entry,
struct dentry *newent; struct dentry *newent;
bool outarg_valid = true; bool outarg_valid = true;
fuse_lock_inode(dir);
err = fuse_lookup_name(dir->i_sb, get_node_id(dir), &entry->d_name, err = fuse_lookup_name(dir->i_sb, get_node_id(dir), &entry->d_name,
&outarg, &inode); &outarg, &inode);
fuse_unlock_inode(dir);
if (err == -ENOENT) { if (err == -ENOENT) {
outarg_valid = false; outarg_valid = false;
err = 0; err = 0;
...@@ -1341,7 +1343,9 @@ static int fuse_readdir(struct file *file, struct dir_context *ctx) ...@@ -1341,7 +1343,9 @@ static int fuse_readdir(struct file *file, struct dir_context *ctx)
fuse_read_fill(req, file, ctx->pos, PAGE_SIZE, fuse_read_fill(req, file, ctx->pos, PAGE_SIZE,
FUSE_READDIR); FUSE_READDIR);
} }
fuse_lock_inode(inode);
fuse_request_send(fc, req); fuse_request_send(fc, req);
fuse_unlock_inode(inode);
nbytes = req->out.args[0].size; nbytes = req->out.args[0].size;
err = req->out.h.error; err = req->out.h.error;
fuse_put_request(fc, req); fuse_put_request(fc, req);
......
...@@ -110,6 +110,9 @@ struct fuse_inode { ...@@ -110,6 +110,9 @@ struct fuse_inode {
/** Miscellaneous bits describing inode state */ /** Miscellaneous bits describing inode state */
unsigned long state; unsigned long state;
/** Lock for serializing lookup and readdir for back compatibility*/
struct mutex mutex;
}; };
/** FUSE inode state bits */ /** FUSE inode state bits */
...@@ -540,6 +543,9 @@ struct fuse_conn { ...@@ -540,6 +543,9 @@ struct fuse_conn {
/** write-back cache policy (default is write-through) */ /** write-back cache policy (default is write-through) */
unsigned writeback_cache:1; unsigned writeback_cache:1;
/** allow parallel lookups and readdir (default is serialized) */
unsigned parallel_dirops:1;
/* /*
* The following bitfields are only for optimization purposes * The following bitfields are only for optimization purposes
* and hence races in setting them will not cause malfunction * and hence races in setting them will not cause malfunction
...@@ -956,4 +962,7 @@ int fuse_do_setattr(struct inode *inode, struct iattr *attr, ...@@ -956,4 +962,7 @@ int fuse_do_setattr(struct inode *inode, struct iattr *attr,
void fuse_set_initialized(struct fuse_conn *fc); void fuse_set_initialized(struct fuse_conn *fc);
void fuse_unlock_inode(struct inode *inode);
void fuse_lock_inode(struct inode *inode);
#endif /* _FS_FUSE_I_H */ #endif /* _FS_FUSE_I_H */
...@@ -97,6 +97,7 @@ static struct inode *fuse_alloc_inode(struct super_block *sb) ...@@ -97,6 +97,7 @@ static struct inode *fuse_alloc_inode(struct super_block *sb)
INIT_LIST_HEAD(&fi->queued_writes); INIT_LIST_HEAD(&fi->queued_writes);
INIT_LIST_HEAD(&fi->writepages); INIT_LIST_HEAD(&fi->writepages);
init_waitqueue_head(&fi->page_waitq); init_waitqueue_head(&fi->page_waitq);
mutex_init(&fi->mutex);
fi->forget = fuse_alloc_forget(); fi->forget = fuse_alloc_forget();
if (!fi->forget) { if (!fi->forget) {
kmem_cache_free(fuse_inode_cachep, inode); kmem_cache_free(fuse_inode_cachep, inode);
...@@ -117,6 +118,7 @@ static void fuse_destroy_inode(struct inode *inode) ...@@ -117,6 +118,7 @@ static void fuse_destroy_inode(struct inode *inode)
struct fuse_inode *fi = get_fuse_inode(inode); struct fuse_inode *fi = get_fuse_inode(inode);
BUG_ON(!list_empty(&fi->write_files)); BUG_ON(!list_empty(&fi->write_files));
BUG_ON(!list_empty(&fi->queued_writes)); BUG_ON(!list_empty(&fi->queued_writes));
mutex_destroy(&fi->mutex);
kfree(fi->forget); kfree(fi->forget);
call_rcu(&inode->i_rcu, fuse_i_callback); call_rcu(&inode->i_rcu, fuse_i_callback);
} }
...@@ -351,6 +353,18 @@ int fuse_reverse_inval_inode(struct super_block *sb, u64 nodeid, ...@@ -351,6 +353,18 @@ int fuse_reverse_inval_inode(struct super_block *sb, u64 nodeid,
return 0; return 0;
} }
void fuse_lock_inode(struct inode *inode)
{
if (!get_fuse_conn(inode)->parallel_dirops)
mutex_lock(&get_fuse_inode(inode)->mutex);
}
void fuse_unlock_inode(struct inode *inode)
{
if (!get_fuse_conn(inode)->parallel_dirops)
mutex_unlock(&get_fuse_inode(inode)->mutex);
}
static void fuse_umount_begin(struct super_block *sb) static void fuse_umount_begin(struct super_block *sb)
{ {
fuse_abort_conn(get_fuse_conn_super(sb)); fuse_abort_conn(get_fuse_conn_super(sb));
...@@ -898,6 +912,8 @@ static void process_init_reply(struct fuse_conn *fc, struct fuse_req *req) ...@@ -898,6 +912,8 @@ static void process_init_reply(struct fuse_conn *fc, struct fuse_req *req)
fc->async_dio = 1; fc->async_dio = 1;
if (arg->flags & FUSE_WRITEBACK_CACHE) if (arg->flags & FUSE_WRITEBACK_CACHE)
fc->writeback_cache = 1; fc->writeback_cache = 1;
if (arg->flags & FUSE_PARALLEL_DIROPS)
fc->parallel_dirops = 1;
if (arg->time_gran && arg->time_gran <= 1000000000) if (arg->time_gran && arg->time_gran <= 1000000000)
fc->sb->s_time_gran = arg->time_gran; fc->sb->s_time_gran = arg->time_gran;
} else { } else {
...@@ -928,7 +944,8 @@ static void fuse_send_init(struct fuse_conn *fc, struct fuse_req *req) ...@@ -928,7 +944,8 @@ static void fuse_send_init(struct fuse_conn *fc, struct fuse_req *req)
FUSE_SPLICE_WRITE | FUSE_SPLICE_MOVE | FUSE_SPLICE_READ | FUSE_SPLICE_WRITE | FUSE_SPLICE_MOVE | FUSE_SPLICE_READ |
FUSE_FLOCK_LOCKS | FUSE_IOCTL_DIR | FUSE_AUTO_INVAL_DATA | FUSE_FLOCK_LOCKS | FUSE_IOCTL_DIR | FUSE_AUTO_INVAL_DATA |
FUSE_DO_READDIRPLUS | FUSE_READDIRPLUS_AUTO | FUSE_ASYNC_DIO | FUSE_DO_READDIRPLUS | FUSE_READDIRPLUS_AUTO | FUSE_ASYNC_DIO |
FUSE_WRITEBACK_CACHE | FUSE_NO_OPEN_SUPPORT; FUSE_WRITEBACK_CACHE | FUSE_NO_OPEN_SUPPORT |
FUSE_PARALLEL_DIROPS;
req->in.h.opcode = FUSE_INIT; req->in.h.opcode = FUSE_INIT;
req->in.numargs = 1; req->in.numargs = 1;
req->in.args[0].size = sizeof(*arg); req->in.args[0].size = sizeof(*arg);
......
...@@ -105,6 +105,9 @@ ...@@ -105,6 +105,9 @@
* *
* 7.24 * 7.24
* - add FUSE_LSEEK for SEEK_HOLE and SEEK_DATA support * - add FUSE_LSEEK for SEEK_HOLE and SEEK_DATA support
*
* 7.25
* - add FUSE_PARALLEL_DIROPS
*/ */
#ifndef _LINUX_FUSE_H #ifndef _LINUX_FUSE_H
...@@ -140,7 +143,7 @@ ...@@ -140,7 +143,7 @@
#define FUSE_KERNEL_VERSION 7 #define FUSE_KERNEL_VERSION 7
/** Minor version number of this interface */ /** Minor version number of this interface */
#define FUSE_KERNEL_MINOR_VERSION 24 #define FUSE_KERNEL_MINOR_VERSION 25
/** The node ID of the root inode */ /** The node ID of the root inode */
#define FUSE_ROOT_ID 1 #define FUSE_ROOT_ID 1
...@@ -234,6 +237,7 @@ struct fuse_file_lock { ...@@ -234,6 +237,7 @@ struct fuse_file_lock {
* FUSE_ASYNC_DIO: asynchronous direct I/O submission * FUSE_ASYNC_DIO: asynchronous direct I/O submission
* FUSE_WRITEBACK_CACHE: use writeback cache for buffered writes * FUSE_WRITEBACK_CACHE: use writeback cache for buffered writes
* FUSE_NO_OPEN_SUPPORT: kernel supports zero-message opens * FUSE_NO_OPEN_SUPPORT: kernel supports zero-message opens
* FUSE_PARALLEL_DIROPS: allow parallel lookups and readdir
*/ */
#define FUSE_ASYNC_READ (1 << 0) #define FUSE_ASYNC_READ (1 << 0)
#define FUSE_POSIX_LOCKS (1 << 1) #define FUSE_POSIX_LOCKS (1 << 1)
...@@ -253,6 +257,7 @@ struct fuse_file_lock { ...@@ -253,6 +257,7 @@ struct fuse_file_lock {
#define FUSE_ASYNC_DIO (1 << 15) #define FUSE_ASYNC_DIO (1 << 15)
#define FUSE_WRITEBACK_CACHE (1 << 16) #define FUSE_WRITEBACK_CACHE (1 << 16)
#define FUSE_NO_OPEN_SUPPORT (1 << 17) #define FUSE_NO_OPEN_SUPPORT (1 << 17)
#define FUSE_PARALLEL_DIROPS (1 << 18)
/** /**
* CUSE INIT request/reply flags * CUSE INIT request/reply flags
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment