Thank you for your contribution! Sashiko AI review found 3 potential issue(s) to consider: - [Medium] tracefs: missing TRACEFS_RO_EVENT_INODE on root in eventfs_ro_fill_super() - [Medium] tracefs: stale TRACEFS_RO_EVENT_INODE retained on slab reuse - [Low] tracefs: kernel-doc function name mismatch on __eventfs_root_lookup() --
commit fbf6ad09ba162cb7562d4fb9830616b2ca868703 Author: Anubhav Shelat <[email protected]> tracefs: Add read-only eventfs filesystem at /sys/kernel/events This commit adds a read-only eventfs pseudo-filesystem mounted at /sys/kernel/events that exposes trace event format and id files with mode 0444 to unprivileged users, reusing the existing eventfs_inode lazy-lookup infrastructure and tracefs inode allocator. > diff --git a/fs/tracefs/event_inode.c b/fs/tracefs/event_inode.c > index 6e3513b13cfa2..b7c3ce3f0d14d 100644 > --- a/fs/tracefs/event_inode.c > +++ b/fs/tracefs/event_inode.c > @@ -16,14 +16,100 @@ > */ > #include <linux/fsnotify.h> > #include <linux/fs.h> > +#include <linux/fs_context.h> > #include <linux/namei.h> > #include <linux/workqueue.h> > #include <linux/security.h> > #include <linux/tracefs.h> > +#include <linux/magic.h> > #include <linux/kref.h> > #include <linux/delay.h> > #include "internal.h" > > +static struct vfsmount *eventfs_ro_mount; > +static int eventfs_ro_mount_count; > + > +static int eventfs_ro_fill_super(struct super_block *sb, struct fs_context > *fc) > +{ > + struct inode *inode; > + struct dentry *root; > + > + sb->s_blocksize = PAGE_SIZE; > + sb->s_blocksize_bits = PAGE_SHIFT; > + sb->s_magic = EVENTFS_SUPER_MAGIC; > + sb->s_op = &eventfs_ro_super_operations; > + sb->s_time_gran = 1; > + sb->s_flags |= SB_RDONLY; > + > + inode = new_inode(sb); > + if (!inode) > + return -ENOMEM; [Severity: Medium] As reported by Sashiko on v4 [1], in eventfs_ro_fill_super(), new_inode(sb) invokes tracefs_alloc_inode() via eventfs_ro_super_operations and links the root tracefs_inode onto the global tracefs_inodes list without setting TRACEFS_RO_EVENT_INODE. TRACEFS_RO_EVENT_INODE is only set on the root inode later in eventfs_create_events_ro_copy(). However, if LOCKDOWN_TRACEFS is active at boot, create_event_toplevel_files() returns -EPERM before calling eventfs_create_events_ro_copy(): e_events = eventfs_create_events_dir("events", parent, events_entries, nr_entries, tr); ... /* Lockdown causes e_events to be NULL*/ if (!e_events) return -EPERM; If eventfs is mounted by userspace (or in the window during eventfs_ro_get_root() before eventfs_create_events_ro_copy() sets ti->flags), can a tracefs remount with uid= or gid= options walk tracefs_inodes in tracefs_apply_options() and overwrite i_uid and i_gid on the read-only eventfs root inode? Should TRACEFS_RO_EVENT_INODE be set on get_tracefs(inode)->flags directly in eventfs_ro_fill_super() when the root inode is allocated? [ ... ] > @@ -327,7 +432,7 @@ static void update_inode_attr(struct inode *inode, > umode_t mode, > } > > static struct inode *eventfs_get_inode(struct dentry *dentry, struct > eventfs_attr *attr, > - umode_t mode, struct eventfs_inode *ei) > + umode_t mode, struct eventfs_inode *ei, > bool ro) > { > struct eventfs_root_inode *rei; > struct eventfs_inode *pei; > @@ -342,6 +447,9 @@ static struct inode *eventfs_get_inode(struct dentry > *dentry, struct eventfs_att > ti->private = ei; > ti->flags |= TRACEFS_EVENT_INODE; > > + if (ro) > + ti->flags |= TRACEFS_RO_EVENT_INODE; [Severity: Medium] When a read-only eventfs inode is evicted and its struct tracefs_inode slab object is later reused for a normal tracefs inode, does the stale TRACEFS_RO_EVENT_INODE flag remain set in ti->flags? Because eventfs_ro_super_operations shares tracefs_inode_cachep, an evicted read-only eventfs inode goes through tracefs_drop_inode(), which only clears TRACEFS_EVENT_INODE: static int tracefs_drop_inode(struct inode *inode) { struct tracefs_inode *ti = get_tracefs(inode); ... ti->flags &= ~TRACEFS_EVENT_INODE; return 1; } The slab constructor init_once() only runs when a new slab is allocated, not when a freed object is reallocated by tracefs_alloc_inode() via alloc_inode_sb(), and neither tracefs_alloc_inode() nor callers such as tracefs_create_file(), __create_dir(), eventfs_create_events_dir(), or eventfs_get_inode(..., ro=false) reset ti->flags. If a slab object previously used for a read-only eventfs inode is reused for a tracefs inode (or the tracefs events directory inode), won't tracefs_apply_options() see the stale TRACEFS_RO_EVENT_INODE bit and skip updating its uid/gid and calling eventfs_remount()? [ ... ] > @@ -521,14 +647,16 @@ lookup_file_dentry(struct dentry *dentry, > * @dir: in which a lookup is being done > * @dentry: file/dir dentry > * @flags: Just passed to simple_lookup() > + * @ro: Denote if this is from a read-only eventfs > * > * Used to create dynamic file/dir with-in @dir, search with-in @ei > * list, if @dentry found go ahead and create the file/dir > */ > > -static struct dentry *eventfs_root_lookup(struct inode *dir, > - struct dentry *dentry, > - unsigned int flags) > +static struct dentry *__eventfs_root_lookup(struct inode *dir, > + struct dentry *dentry, > + unsigned int flags, > + bool ro) [Severity: Low] As reported by Sashiko on v5 [2], this isn't a bug, but since eventfs_root_lookup() was renamed to __eventfs_root_lookup(), should the function name in the kernel-doc comment header above it also be updated to __eventfs_root_lookup to avoid a W=1 kernel-doc warning? [1] https://lore.kernel.org/r/[email protected] [2] https://lore.kernel.org/r/[email protected] -- Sashiko AI review ยท https://sashiko.dev/#/patchset/[email protected]?part=1
