On 25/07/08 09:27PM, Darrick J. Wong wrote:
> On Thu, Jul 03, 2025 at 01:50:26PM -0500, John Groves wrote:
> > Upon completion of an OPEN, if we're in famfs-mode we do a GET_FMAP to
> > retrieve and cache up the file-to-dax map in the kernel. If this
> > succeeds, read/write/mmap are resolved direct-to-dax with no upcalls.
> >
> > GET_FMAP has a variable-size response payload, and the allocated size
> > is sent in the in_args[0].size field. If the fmap would overflow the
> > message, the fuse server sends a reply of size 'sizeof(uint32_t)' which
> > specifies the size of the fmap message. Then the kernel can realloc a
> > large enough buffer and try again.
> >
> > Signed-off-by: John Groves <john@xxxxxxxxxx>
> > ---
> > fs/fuse/file.c | 84 +++++++++++++++++++++++++++++++++++++++
> > fs/fuse/fuse_i.h | 36 ++++++++++++++++-
> > fs/fuse/inode.c | 19 +++++++--
> > fs/fuse/iomode.c | 2 +-
> > include/uapi/linux/fuse.h | 18 +++++++++
> > 5 files changed, 154 insertions(+), 5 deletions(-)
> >
> > diff --git a/fs/fuse/file.c b/fs/fuse/file.c
> > index 93b82660f0c8..8616fb0a6d61 100644
> > --- a/fs/fuse/file.c
> > +++ b/fs/fuse/file.c
> > @@ -230,6 +230,77 @@ static void fuse_truncate_update_attr(struct inode *inode, struct file *file)
> > fuse_invalidate_attr_mask(inode, FUSE_STATX_MODSIZE);
> > }
> >
> > +#if IS_ENABLED(CONFIG_FUSE_FAMFS_DAX)
> > +
> > +#define FMAP_BUFSIZE 4096
>
> PAGE_SIZE ?
Like it. Queued to -next
>
> > +
> > +static int
> > +fuse_get_fmap(struct fuse_mount *fm, struct inode *inode, u64 nodeid)
> > +{
> > + struct fuse_get_fmap_in inarg = { 0 };
> > + size_t fmap_bufsize = FMAP_BUFSIZE;
> > + ssize_t fmap_size;
> > + int retries = 1;
> > + void *fmap_buf;
> > + int rc;
> > +
> > + FUSE_ARGS(args);
> > +
> > + fmap_buf = kcalloc(1, FMAP_BUFSIZE, GFP_KERNEL);
> > + if (!fmap_buf)
> > + return -EIO;
> > +
> > + retry_once:
> > + inarg.size = fmap_bufsize;
> > +
> > + args.opcode = FUSE_GET_FMAP;
> > + args.nodeid = nodeid;
> > +
> > + args.in_numargs = 1;
> > + args.in_args[0].size = sizeof(inarg);
> > + args.in_args[0].value = &inarg;
> > +
> > + /* Variable-sized output buffer
> > + * this causes fuse_simple_request() to return the size of the
> > + * output payload
> > + */
> > + args.out_argvar = true;
> > + args.out_numargs = 1;
> > + args.out_args[0].size = fmap_bufsize;
> > + args.out_args[0].value = fmap_buf;
> > +
> > + /* Send GET_FMAP command */
> > + rc = fuse_simple_request(fm, &args);
> > + if (rc < 0) {
> > + pr_err("%s: err=%d from fuse_simple_request()\n",
> > + __func__, rc);
> > + return rc;
> > + }
> > + fmap_size = rc;
> > +
> > + if (retries && fmap_size == sizeof(uint32_t)) {
> > + /* fmap size exceeded fmap_bufsize;
> > + * actual fmap size returned in fmap_buf;
> > + * realloc and retry once
> > + */
> > + fmap_bufsize = *((uint32_t *)fmap_buf);
> > +
> > + --retries;
> > + kfree(fmap_buf);
> > + fmap_buf = kcalloc(1, fmap_bufsize, GFP_KERNEL);
> > + if (!fmap_buf)
> > + return -EIO;
> > +
> > + goto retry_once;
> > + }
> > +
> > + /* Will call famfs_file_init_dax() when that gets added */
>
> Hard to say what this does without looking further down in the patchset.
> :)
New comment:
/* We retrieved the "fmap" (the file's map to memory), but
* we haven't used it yet. A call to famfs_file_init_dax() will be added
* here in a subsequent patch, when we add the ability to attach
* fmaps to files.
*/
>
> > + kfree(fmap_buf);
> > + return 0;
> > +}
> > +#endif
> > +
> > static int fuse_open(struct inode *inode, struct file *file)
> > {
> > struct fuse_mount *fm = get_fuse_mount(inode);
> > @@ -263,6 +334,19 @@ static int fuse_open(struct inode *inode, struct file *file)
> >
> > err = fuse_do_open(fm, get_node_id(inode), file, false);
> > if (!err) {
> > +#if IS_ENABLED(CONFIG_FUSE_FAMFS_DAX)
> > + if (fm->fc->famfs_iomap) {
> > + if (S_ISREG(inode->i_mode)) {
>
> /me wonders if you want to turn this into a dumb helper to reduce the
> indenting levels?
>
> #if IS_ENABLED(CONFIG_FUSE_FAMFS_DAX)
> static inline bool fuse_is_famfs_file(struct inode *inode)
> {
> return fm->fc->famfs_iomap && S_ISREG(inode->i_mode);
> }
> #else
> # define fuse_is_famfs_file(...) (false)
> #endif
>
> if (!err) {
> if (fuse_is_famfs_file(inode)) {
> rc = fuse_get_fmap(fm, inode);
> ...
> }
> }
>
I've already refactored helpers and simplified this logic in the -next
branch, including losing the conditrional code here in file.c:
if (!err) {
if ((fm->fc->famfs_iomap) && (S_ISREG(inode->i_mode))) {
int rc;
/* Get the famfs fmap */
rc = fuse_get_fmap(fm, inode);
...
}
...
}
So I think it's quite a bit cleaner... will send out an updated patch
pretty soon (probably next week, without the poisoned page fixes yet).
> > + int rc;
> > + /* Get the famfs fmap */
> > + rc = fuse_get_fmap(fm, inode,
> > + get_node_id(inode));
>
> Just get_node_id inside fuse_get_fmap to reduce the parameter count.
Done, thanks
>
> > + if (rc)
> > + pr_err("%s: fuse_get_fmap err=%d\n",
> > + __func__, rc);
> > + }
> > + }
> > +#endif
> > ff = file->private_data;
> > err = fuse_finish_open(inode, file);
> > if (err)
> > diff --git a/fs/fuse/fuse_i.h b/fs/fuse/fuse_i.h
> > index f4ee61046578..e01d6e5c6e93 100644
> > --- a/fs/fuse/fuse_i.h
> > +++ b/fs/fuse/fuse_i.h
> > @@ -193,6 +193,10 @@ struct fuse_inode {
> > /** Reference to backing file in passthrough mode */
> > struct fuse_backing *fb;
> > #endif
> > +
> > +#if IS_ENABLED(CONFIG_FUSE_FAMFS_DAX)
> > + void *famfs_meta;
> > +#endif
>
> What gets stored in here?
Explanatory comment added:
/* Pointer to the file's famfs metadata. Primary content is the
* in-memory version of the fmap - the map from file's offset range
* to DAX memory
*/
>
> > };
> >
> > /** FUSE inode state bits */
> > @@ -945,6 +949,8 @@ struct fuse_conn {
> > #endif
> >
> > #if IS_ENABLED(CONFIG_FUSE_FAMFS_DAX)
> > + struct rw_semaphore famfs_devlist_sem;
> > + struct famfs_dax_devlist *dax_devlist;
> > char *shadow;
> > #endif
> > };
> > @@ -1435,11 +1441,14 @@ void fuse_free_conn(struct fuse_conn *fc);
> >
> > /* dax.c */
> >
> > +static inline int fuse_file_famfs(struct fuse_inode *fi); /* forward */
> > +
> > /* This macro is used by virtio_fs, but now it also needs to filter for
> > * "not famfs"
> > */
> > #define FUSE_IS_VIRTIO_DAX(fuse_inode) (IS_ENABLED(CONFIG_FUSE_DAX) \
> > - && IS_DAX(&fuse_inode->inode))
> > + && IS_DAX(&fuse_inode->inode) \
> > + && !fuse_file_famfs(fuse_inode))
> >
> > ssize_t fuse_dax_read_iter(struct kiocb *iocb, struct iov_iter *to);
> > ssize_t fuse_dax_write_iter(struct kiocb *iocb, struct iov_iter *from);
> > @@ -1550,4 +1559,29 @@ extern void fuse_sysctl_unregister(void);
> > #define fuse_sysctl_unregister() do { } while (0)
> > #endif /* CONFIG_SYSCTL */
> >
> > +/* famfs.c */
> > +static inline struct fuse_backing *famfs_meta_set(struct fuse_inode *fi,
> > + void *meta)
> > +{
> > +#if IS_ENABLED(CONFIG_FUSE_FAMFS_DAX)
> > + return xchg(&fi->famfs_meta, meta);
> > +#else
> > + return NULL;
> > +#endif
> > +}
> > +
> > +static inline void famfs_meta_free(struct fuse_inode *fi)
> > +{
> > + /* Stub wil be connected in a subsequent commit */
> > +}
> > +
> > +static inline int fuse_file_famfs(struct fuse_inode *fi)
> > +{
> > +#if IS_ENABLED(CONFIG_FUSE_FAMFS_DAX)
> > + return (READ_ONCE(fi->famfs_meta) != NULL);
> > +#else
> > + return 0;
> > +#endif
> > +}
>
> ...or maybe this is the predicate you want to see if you really need to
> fmapping related stuff?
>
> > +
> > #endif /* _FS_FUSE_I_H */
> > diff --git a/fs/fuse/inode.c b/fs/fuse/inode.c
> > index a7e1cf8257b0..b071d16f7d04 100644
> > --- a/fs/fuse/inode.c
> > +++ b/fs/fuse/inode.c
> > @@ -117,6 +117,9 @@ static struct inode *fuse_alloc_inode(struct super_block *sb)
> > if (IS_ENABLED(CONFIG_FUSE_PASSTHROUGH))
> > fuse_inode_backing_set(fi, NULL);
> >
> > + if (IS_ENABLED(CONFIG_FUSE_FAMFS_DAX))
> > + famfs_meta_set(fi, NULL);
> > +
> > return &fi->inode;
> >
> > out_free_forget:
> > @@ -138,6 +141,13 @@ static void fuse_free_inode(struct inode *inode)
> > if (IS_ENABLED(CONFIG_FUSE_PASSTHROUGH))
> > fuse_backing_put(fuse_inode_backing(fi));
> >
> > +#if IS_ENABLED(CONFIG_FUSE_FAMFS_DAX)
> > + if (S_ISREG(inode->i_mode) && fi->famfs_meta) {
> > + famfs_meta_free(fi);
> > + famfs_meta_set(fi, NULL);
>
> _free should null out the pointer, no?
Good point - will do
<snip>
Thanks Darrick!
John
