| Message ID | 20240315-dir-deleg-v1-8-a1d6209a3654@kernel.org (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | vfs, nfsd, nfs: implement directory delegations | expand |
> int vfs_mknod(struct mnt_idmap *, struct inode *, struct dentry *, > - umode_t, dev_t); > + umode_t, dev_t, struct inode **); So we will have at least the following helpers with an additional delegated inode argument. vfs_unlink() vfs_link() notify_change() vfs_create() vfs_mknod() vfs_mkdir() vfs_rmdir() From looking at callers all these helpers will be called with non-NULL delegated inode argument in vfs only. Unless it is generally conceivable that other callers will want to pass a non-NULL inode argument over time it might make more sense to add vfs_<operation>_delegated() or __vfs_<operation>() and make vfs_mknod() and friends exported wrappers around it. I mean it's a matter of preference ultimately but this seems cleaner to me. So at least for the new ones we should consider it. Would also make the patch smaller. > int vfs_symlink(struct mnt_idmap *, struct inode *, > struct dentry *, const char *); > int vfs_link(struct dentry *, struct mnt_idmap *, struct inode *, > @@ -1879,7 +1879,7 @@ static inline int vfs_whiteout(struct mnt_idmap *idmap, > struct inode *dir, struct dentry *dentry) > { > return vfs_mknod(idmap, dir, dentry, S_IFCHR | WHITEOUT_MODE, > - WHITEOUT_DEV); > + WHITEOUT_DEV, NULL); > } > > struct file *kernel_tmpfile_open(struct mnt_idmap *idmap, > diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c > index 0748e7ea5210..34fbcc90c984 100644 > --- a/net/unix/af_unix.c > +++ b/net/unix/af_unix.c > @@ -1227,7 +1227,7 @@ static int unix_bind_bsd(struct sock *sk, struct sockaddr_un *sunaddr, > idmap = mnt_idmap(parent.mnt); > err = security_path_mknod(&parent, dentry, mode, 0); > if (!err) > - err = vfs_mknod(idmap, d_inode(parent.dentry), dentry, mode, 0); > + err = vfs_mknod(idmap, d_inode(parent.dentry), dentry, mode, 0, NULL); > if (err) > goto out_path; > err = mutex_lock_interruptible(&u->bindlock); > > -- > 2.44.0 >
On Wed, 2024-03-20 at 14:42 +0100, Christian Brauner wrote: > > int vfs_mknod(struct mnt_idmap *, struct inode *, struct dentry *, > > - umode_t, dev_t); > > + umode_t, dev_t, struct inode **); > > So we will have at least the following helpers with an additional > delegated inode argument. > > vfs_unlink() > vfs_link() > notify_change() > vfs_create() > vfs_mknod() > vfs_mkdir() > vfs_rmdir() > > From looking at callers all these helpers will be called with non-NULL > delegated inode argument in vfs only. Unless it is generally conceivable > that other callers will want to pass a non-NULL inode argument over time > it might make more sense to add vfs_<operation>_delegated() or > __vfs_<operation>() and make vfs_mknod() and friends exported wrappers > around it. > > I mean it's a matter of preference ultimately but this seems cleaner to > me. So at least for the new ones we should consider it. Would also make > the patch smaller. > Good suggestion. I just respun along those lines and it's a lot cleaner. I'm still testing it but here is the new diffstat. It's a little larger actually, but it keeps the changes more confined to namei.c: jlayton@tleilax:~/git/linux$ git diff master --stat fs/locks.c | 12 +++- fs/namei.c | 227 ++++++++++++++++++++++++++++++++++++++++++++++-------------------- fs/nfs/delegation.c | 5 ++ fs/nfs/dir.c | 20 ++++++ fs/nfs/internal.h | 2 +- fs/nfs/nfs4file.c | 2 + fs/nfs/nfs4proc.c | 62 +++++++++++++++++- fs/nfs/nfs4trace.h | 104 ++++++++++++++++++++++++++++++ fs/nfs/nfs4xdr.c | 136 +++++++++++++++++++++++++++++++++++++++ fs/nfs/nfstrace.h | 8 ++- fs/nfsd/filecache.c | 37 +++++++++-- fs/nfsd/filecache.h | 2 + fs/nfsd/nfs4proc.c | 48 ++++++++++++++ fs/nfsd/nfs4state.c | 113 ++++++++++++++++++++++++++++++++- fs/nfsd/nfs4xdr.c | 91 ++++++++++++++++++++++++++- fs/nfsd/state.h | 5 ++ fs/nfsd/vfs.c | 5 +- fs/nfsd/vfs.h | 2 +- fs/nfsd/xdr4.h | 19 ++++++ fs/smb/client/cifsfs.c | 3 + include/linux/filelock.h | 14 +++++ include/linux/nfs4.h | 7 +++ include/linux/nfs_fs.h | 1 + include/linux/nfs_fs_sb.h | 1 + include/linux/nfs_xdr.h | 2 + 25 files changed, 838 insertions(+), 90 deletions(-)
On Wed, Mar 20, 2024 at 04:12:29PM -0400, Jeff Layton wrote: > On Wed, 2024-03-20 at 14:42 +0100, Christian Brauner wrote: > > > int vfs_mknod(struct mnt_idmap *, struct inode *, struct dentry *, > > > - umode_t, dev_t); > > > + umode_t, dev_t, struct inode **); > > > > So we will have at least the following helpers with an additional > > delegated inode argument. > > > > vfs_unlink() > > vfs_link() > > notify_change() > > vfs_create() > > vfs_mknod() > > vfs_mkdir() > > vfs_rmdir() > > > > From looking at callers all these helpers will be called with non-NULL > > delegated inode argument in vfs only. Unless it is generally conceivable > > that other callers will want to pass a non-NULL inode argument over time > > it might make more sense to add vfs_<operation>_delegated() or > > __vfs_<operation>() and make vfs_mknod() and friends exported wrappers > > around it. > > > > I mean it's a matter of preference ultimately but this seems cleaner to > > me. So at least for the new ones we should consider it. Would also make > > the patch smaller. > > > > Good suggestion. I just respun along those lines and it's a lot cleaner. > I'm still testing it but here is the new diffstat. It's a little larger > actually, but it keeps the changes more confined to namei.c: Sounds good to me!
diff --git a/drivers/base/devtmpfs.c b/drivers/base/devtmpfs.c index c00126796f79..8c0a872e3165 100644 --- a/drivers/base/devtmpfs.c +++ b/drivers/base/devtmpfs.c @@ -217,7 +217,7 @@ static int handle_create(const char *nodename, umode_t mode, kuid_t uid, return PTR_ERR(dentry); err = vfs_mknod(&nop_mnt_idmap, d_inode(path.dentry), dentry, mode, - dev->devt); + dev->devt, NULL); if (!err) { struct iattr newattrs; diff --git a/fs/ecryptfs/inode.c b/fs/ecryptfs/inode.c index a99b1e264c46..c6442b8caa2f 100644 --- a/fs/ecryptfs/inode.c +++ b/fs/ecryptfs/inode.c @@ -566,7 +566,7 @@ ecryptfs_mknod(struct mnt_idmap *idmap, struct inode *dir, rc = lock_parent(dentry, &lower_dentry, &lower_dir); if (!rc) rc = vfs_mknod(&nop_mnt_idmap, lower_dir, - lower_dentry, mode, dev); + lower_dentry, mode, dev, NULL); if (rc || d_really_is_negative(lower_dentry)) goto out; rc = ecryptfs_interpose(lower_dentry, dentry, dir->i_sb); diff --git a/fs/init.c b/fs/init.c index 325c9e4d9b20..99c6b413adad 100644 --- a/fs/init.c +++ b/fs/init.c @@ -157,7 +157,7 @@ int __init init_mknod(const char *filename, umode_t mode, unsigned int dev) error = security_path_mknod(&path, dentry, mode, dev); if (!error) error = vfs_mknod(mnt_idmap(path.mnt), path.dentry->d_inode, - dentry, mode, new_decode_dev(dev)); + dentry, mode, new_decode_dev(dev), NULL); done_path_create(&path, dentry); return error; } diff --git a/fs/namei.c b/fs/namei.c index 01e04cf155eb..a185974c1a55 100644 --- a/fs/namei.c +++ b/fs/namei.c @@ -3996,7 +3996,8 @@ EXPORT_SYMBOL(user_path_create); * raw inode simply pass @nop_mnt_idmap. */ int vfs_mknod(struct mnt_idmap *idmap, struct inode *dir, - struct dentry *dentry, umode_t mode, dev_t dev) + struct dentry *dentry, umode_t mode, dev_t dev, + struct inode **delegated_inode) { bool is_whiteout = S_ISCHR(mode) && dev == WHITEOUT_DEV; int error = may_create(idmap, dir, dentry); @@ -4020,6 +4021,10 @@ int vfs_mknod(struct mnt_idmap *idmap, struct inode *dir, if (error) return error; + error = try_break_deleg(dir, delegated_inode); + if (error) + return error; + error = dir->i_op->mknod(idmap, dir, dentry, mode, dev); if (!error) fsnotify_create(dir, dentry); @@ -4078,11 +4083,11 @@ static int do_mknodat(int dfd, struct filename *name, umode_t mode, break; case S_IFCHR: case S_IFBLK: error = vfs_mknod(idmap, path.dentry->d_inode, - dentry, mode, new_decode_dev(dev)); + dentry, mode, new_decode_dev(dev), &delegated_inode); break; case S_IFIFO: case S_IFSOCK: error = vfs_mknod(idmap, path.dentry->d_inode, - dentry, mode, 0); + dentry, mode, 0, &delegated_inode); break; } out2: diff --git a/fs/nfsd/vfs.c b/fs/nfsd/vfs.c index 47b8ab1d4b17..fe088e7c49c8 100644 --- a/fs/nfsd/vfs.c +++ b/fs/nfsd/vfs.c @@ -1525,7 +1525,7 @@ nfsd_create_locked(struct svc_rqst *rqstp, struct svc_fh *fhp, case S_IFIFO: case S_IFSOCK: host_err = vfs_mknod(&nop_mnt_idmap, dirp, dchild, - iap->ia_mode, rdev); + iap->ia_mode, rdev, NULL); break; default: printk(KERN_WARNING "nfsd: bad file type %o in nfsd_create\n", diff --git a/fs/overlayfs/overlayfs.h b/fs/overlayfs/overlayfs.h index be2518e6da95..26cdef6c3579 100644 --- a/fs/overlayfs/overlayfs.h +++ b/fs/overlayfs/overlayfs.h @@ -251,7 +251,7 @@ static inline int ovl_do_mknod(struct ovl_fs *ofs, struct inode *dir, struct dentry *dentry, umode_t mode, dev_t dev) { - int err = vfs_mknod(ovl_upper_mnt_idmap(ofs), dir, dentry, mode, dev); + int err = vfs_mknod(ovl_upper_mnt_idmap(ofs), dir, dentry, mode, dev, NULL); pr_debug("mknod(%pd2, 0%o, 0%o) = %i\n", dentry, mode, dev, err); return err; diff --git a/include/linux/fs.h b/include/linux/fs.h index 8fb4101fea49..4b396c9a7a84 100644 --- a/include/linux/fs.h +++ b/include/linux/fs.h @@ -1841,7 +1841,7 @@ int vfs_create(struct mnt_idmap *, struct inode *, int vfs_mkdir(struct mnt_idmap *, struct inode *, struct dentry *, umode_t, struct inode **); int vfs_mknod(struct mnt_idmap *, struct inode *, struct dentry *, - umode_t, dev_t); + umode_t, dev_t, struct inode **); int vfs_symlink(struct mnt_idmap *, struct inode *, struct dentry *, const char *); int vfs_link(struct dentry *, struct mnt_idmap *, struct inode *, @@ -1879,7 +1879,7 @@ static inline int vfs_whiteout(struct mnt_idmap *idmap, struct inode *dir, struct dentry *dentry) { return vfs_mknod(idmap, dir, dentry, S_IFCHR | WHITEOUT_MODE, - WHITEOUT_DEV); + WHITEOUT_DEV, NULL); } struct file *kernel_tmpfile_open(struct mnt_idmap *idmap, diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c index 0748e7ea5210..34fbcc90c984 100644 --- a/net/unix/af_unix.c +++ b/net/unix/af_unix.c @@ -1227,7 +1227,7 @@ static int unix_bind_bsd(struct sock *sk, struct sockaddr_un *sunaddr, idmap = mnt_idmap(parent.mnt); err = security_path_mknod(&parent, dentry, mode, 0); if (!err) - err = vfs_mknod(idmap, d_inode(parent.dentry), dentry, mode, 0); + err = vfs_mknod(idmap, d_inode(parent.dentry), dentry, mode, 0, NULL); if (err) goto out_path; err = mutex_lock_interruptible(&u->bindlock);
In order to add directory delegation support, we need to break delegations on the parent whenever there is going to be a change in the directory. Add a new delegated_inode parameter to vfs_mknod. Most callers will set that to NULL, but do_mknodat can use that to synchronously wait for the delegation break to complete. Signed-off-by: Jeff Layton <jlayton@kernel.org> --- drivers/base/devtmpfs.c | 2 +- fs/ecryptfs/inode.c | 2 +- fs/init.c | 2 +- fs/namei.c | 11 ++++++++--- fs/nfsd/vfs.c | 2 +- fs/overlayfs/overlayfs.h | 2 +- include/linux/fs.h | 4 ++-- net/unix/af_unix.c | 2 +- 8 files changed, 16 insertions(+), 11 deletions(-)