From 72d0d248ca8232dbd30d35b42d0d86e39b3e322b Mon Sep 17 00:00:00 2001 From: Brian Foster Date: Mon, 16 Jul 2012 15:23:48 -0400 Subject: fuse: add FUSE_AUTO_INVAL_DATA init flag FUSE_AUTO_INVAL_DATA is provided to enable updated/auto cache invalidation logic. Signed-off-by: Brian Foster Signed-off-by: Miklos Szeredi --- fs/fuse/fuse_i.h | 3 +++ fs/fuse/inode.c | 4 +++- 2 files changed, 6 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/fuse/fuse_i.h b/fs/fuse/fuse_i.h index 771fb6322c0..e24dd74e306 100644 --- a/fs/fuse/fuse_i.h +++ b/fs/fuse/fuse_i.h @@ -484,6 +484,9 @@ struct fuse_conn { /** Is fallocate not implemented by fs? */ unsigned no_fallocate:1; + /** Use enhanced/automatic page cache invalidation. */ + unsigned auto_inval_data:1; + /** The number of requests waiting for completion */ atomic_t num_waiting; diff --git a/fs/fuse/inode.c b/fs/fuse/inode.c index 1cd61652018..dd37ee291b8 100644 --- a/fs/fuse/inode.c +++ b/fs/fuse/inode.c @@ -834,6 +834,8 @@ static void process_init_reply(struct fuse_conn *fc, struct fuse_req *req) fc->big_writes = 1; if (arg->flags & FUSE_DONT_MASK) fc->dont_mask = 1; + if (arg->flags & FUSE_AUTO_INVAL_DATA) + fc->auto_inval_data = 1; } else { ra_pages = fc->max_read / PAGE_CACHE_SIZE; fc->no_lock = 1; @@ -859,7 +861,7 @@ static void fuse_send_init(struct fuse_conn *fc, struct fuse_req *req) arg->max_readahead = fc->bdi.ra_pages * PAGE_CACHE_SIZE; arg->flags |= FUSE_ASYNC_READ | FUSE_POSIX_LOCKS | FUSE_ATOMIC_O_TRUNC | FUSE_EXPORT_SUPPORT | FUSE_BIG_WRITES | FUSE_DONT_MASK | - FUSE_FLOCK_LOCKS; + FUSE_FLOCK_LOCKS | FUSE_AUTO_INVAL_DATA; req->in.h.opcode = FUSE_INIT; req->in.numargs = 1; req->in.args[0].size = sizeof(*arg); -- cgit v1.2.3-70-g09d2 From eed2179efe1aac145bf6d54b925b750976380fa6 Mon Sep 17 00:00:00 2001 From: Brian Foster Date: Mon, 16 Jul 2012 15:23:49 -0400 Subject: fuse: invalidate inode mapping if mtime changes We currently invalidate the inode address space mapping if the file size changes unexpectedly. In the case of a fuse network filesystem, a portion of a file could be overwritten remotely without changing the file size. Compare the old mtime as well to detect this condition and invalidate the mapping if the file has been updated. The original logic (to ignore changes in mtime) is preserved unless the client specifies FUSE_AUTO_INVAL_DATA on init. Signed-off-by: Brian Foster Signed-off-by: Miklos Szeredi --- fs/fuse/inode.c | 27 ++++++++++++++++++++++++--- 1 file changed, 24 insertions(+), 3 deletions(-) (limited to 'fs') diff --git a/fs/fuse/inode.c b/fs/fuse/inode.c index dd37ee291b8..dd4401650b4 100644 --- a/fs/fuse/inode.c +++ b/fs/fuse/inode.c @@ -197,6 +197,7 @@ void fuse_change_attributes(struct inode *inode, struct fuse_attr *attr, struct fuse_conn *fc = get_fuse_conn(inode); struct fuse_inode *fi = get_fuse_inode(inode); loff_t oldsize; + struct timespec old_mtime; spin_lock(&fc->lock); if (attr_version != 0 && fi->attr_version > attr_version) { @@ -204,15 +205,35 @@ void fuse_change_attributes(struct inode *inode, struct fuse_attr *attr, return; } + old_mtime = inode->i_mtime; fuse_change_attributes_common(inode, attr, attr_valid); oldsize = inode->i_size; i_size_write(inode, attr->size); spin_unlock(&fc->lock); - if (S_ISREG(inode->i_mode) && oldsize != attr->size) { - truncate_pagecache(inode, oldsize, attr->size); - invalidate_inode_pages2(inode->i_mapping); + if (S_ISREG(inode->i_mode)) { + bool inval = false; + + if (oldsize != attr->size) { + truncate_pagecache(inode, oldsize, attr->size); + inval = true; + } else if (fc->auto_inval_data) { + struct timespec new_mtime = { + .tv_sec = attr->mtime, + .tv_nsec = attr->mtimensec, + }; + + /* + * Auto inval mode also checks and invalidates if mtime + * has changed. + */ + if (!timespec_equal(&old_mtime, &new_mtime)) + inval = true; + } + + if (inval) + invalidate_inode_pages2(inode->i_mapping); } } -- cgit v1.2.3-70-g09d2 From a8894274a3581125fb311eabfc97cd0123740c5e Mon Sep 17 00:00:00 2001 From: Brian Foster Date: Mon, 16 Jul 2012 15:23:50 -0400 Subject: fuse: update attributes on aio_read A fuse-based network filesystem might allow for the inode and/or file data to change unexpectedly. A local client that opens and repeatedly reads a file might never pick up on such changes and indefinitely return stale data. Always invoke fuse_update_attributes() in the read path to cause an attr revalidation when the attributes expire. This leads to a page cache invalidation if necessary and ensures fuse issues new read requests to the fuse client. The original logic (reval only on reads beyond EOF) is preserved unless the client specifies FUSE_AUTO_INVAL_DATA on init. Signed-off-by: Brian Foster Signed-off-by: Miklos Szeredi --- fs/fuse/file.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-) (limited to 'fs') diff --git a/fs/fuse/file.c b/fs/fuse/file.c index b321a688cde..5800101e5ce 100644 --- a/fs/fuse/file.c +++ b/fs/fuse/file.c @@ -703,13 +703,16 @@ static ssize_t fuse_file_aio_read(struct kiocb *iocb, const struct iovec *iov, unsigned long nr_segs, loff_t pos) { struct inode *inode = iocb->ki_filp->f_mapping->host; + struct fuse_conn *fc = get_fuse_conn(inode); - if (pos + iov_length(iov, nr_segs) > i_size_read(inode)) { + /* + * In auto invalidate mode, always update attributes on read. + * Otherwise, only update if we attempt to read past EOF (to ensure + * i_size is up to date). + */ + if (fc->auto_inval_data || + (pos + iov_length(iov, nr_segs) > i_size_read(inode))) { int err; - /* - * If trying to read past EOF, make sure the i_size - * attribute is up-to-date. - */ err = fuse_update_attributes(inode, NULL, iocb->ki_filp, NULL); if (err) return err; -- cgit v1.2.3-70-g09d2 From 69fe05c90ed58aac956dccb9e6d3a325fb3b8767 Mon Sep 17 00:00:00 2001 From: Miklos Szeredi Date: Wed, 18 Jul 2012 16:09:40 +0200 Subject: fuse: add missing INIT flags Add missing flags that userspace derived from the protocol version number. This makes the protocol more flexible. Signed-off-by: Miklos Szeredi --- fs/fuse/inode.c | 3 ++- include/linux/fuse.h | 8 ++++++++ 2 files changed, 10 insertions(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/fuse/inode.c b/fs/fuse/inode.c index dd4401650b4..ce0a2838ccd 100644 --- a/fs/fuse/inode.c +++ b/fs/fuse/inode.c @@ -882,7 +882,8 @@ static void fuse_send_init(struct fuse_conn *fc, struct fuse_req *req) arg->max_readahead = fc->bdi.ra_pages * PAGE_CACHE_SIZE; arg->flags |= FUSE_ASYNC_READ | FUSE_POSIX_LOCKS | FUSE_ATOMIC_O_TRUNC | FUSE_EXPORT_SUPPORT | FUSE_BIG_WRITES | FUSE_DONT_MASK | - FUSE_FLOCK_LOCKS | FUSE_AUTO_INVAL_DATA; + FUSE_SPLICE_WRITE | FUSE_SPLICE_MOVE | FUSE_SPLICE_READ | + FUSE_FLOCK_LOCKS | FUSE_IOCTL_DIR | FUSE_AUTO_INVAL_DATA; req->in.h.opcode = FUSE_INIT; req->in.numargs = 1; req->in.args[0].size = sizeof(*arg); diff --git a/include/linux/fuse.h b/include/linux/fuse.h index e4a9d2af9aa..6455c5b64c2 100644 --- a/include/linux/fuse.h +++ b/include/linux/fuse.h @@ -169,7 +169,11 @@ struct fuse_file_lock { * FUSE_POSIX_LOCKS: remote locking for POSIX file locks * FUSE_EXPORT_SUPPORT: filesystem handles lookups of "." and ".." * FUSE_DONT_MASK: don't apply umask to file mode on create operations + * FUSE_SPLICE_WRITE: kernel supports splice write on the device + * FUSE_SPLICE_MOVE: kernel supports splice move on the device + * FUSE_SPLICE_READ: kernel supports splice read on the device * FUSE_FLOCK_LOCKS: remote locking for BSD style file locks + * FUSE_HAS_IOCTL_DIR: kernel supports ioctl on directories * FUSE_AUTO_INVAL_DATA: automatically invalidate cached pages */ #define FUSE_ASYNC_READ (1 << 0) @@ -179,7 +183,11 @@ struct fuse_file_lock { #define FUSE_EXPORT_SUPPORT (1 << 4) #define FUSE_BIG_WRITES (1 << 5) #define FUSE_DONT_MASK (1 << 6) +#define FUSE_SPLICE_WRITE (1 << 7) +#define FUSE_SPLICE_MOVE (1 << 8) +#define FUSE_SPLICE_READ (1 << 9) #define FUSE_FLOCK_LOCKS (1 << 10) +#define FUSE_HAS_IOCTL_DIR (1 << 11) #define FUSE_AUTO_INVAL_DATA (1 << 12) /** -- cgit v1.2.3-70-g09d2 From fb6ccff667712c46b4501b920ea73a326e49626a Mon Sep 17 00:00:00 2001 From: Zach Brown Date: Tue, 24 Jul 2012 12:10:11 -0700 Subject: fuse: verify all ioctl retry iov elements Commit 7572777eef78ebdee1ecb7c258c0ef94d35bad16 attempted to verify that the total iovec from the client doesn't overflow iov_length() but it only checked the first element. The iovec could still overflow by starting with a small element. The obvious fix is to check all the elements. The overflow case doesn't look dangerous to the kernel as the copy is limited by the length after the overflow. This fix restores the intention of returning an error instead of successfully copying less than the iovec represented. I found this by code inspection. I built it but don't have a test case. I'm cc:ing stable because the initial commit did as well. Signed-off-by: Zach Brown Signed-off-by: Miklos Szeredi CC: [2.6.37+] --- fs/fuse/file.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'fs') diff --git a/fs/fuse/file.c b/fs/fuse/file.c index 5800101e5ce..2eed3acfb6a 100644 --- a/fs/fuse/file.c +++ b/fs/fuse/file.c @@ -1703,7 +1703,7 @@ static int fuse_verify_ioctl_iov(struct iovec *iov, size_t count) size_t n; u32 max = FUSE_MAX_PAGES_PER_REQ << PAGE_SHIFT; - for (n = 0; n < count; n++) { + for (n = 0; n < count; n++, iov++) { if (iov->iov_len > (size_t) max) return -ENOMEM; max -= iov->iov_len; -- cgit v1.2.3-70-g09d2