Skip to contentSkip to conversation

tracing: Two more small fixes for v6.7-rc8

PATCH New Steven Rostedt 1 participant 3 messages Jan 3, 2024

Subsystems (2)
lore
Share & tools
lore

Steven Rostedt

author maintainer
#0 lore raw

cover tracing: Two more small fixes for v6.7-rc8

tracing fixes for v6.7-rc8:

- Fix a NULL kernel dereference in set_gid() on tracefs mounting. When tracefs is mounted with "gid=1000", it will update the existing dentries to have the new gid. The tracefs_inode which is retrieved by a container_of(dentry->d_inode) has flags to see if the inode belongs to the eventfs system.

The issue that was fixed was if getdents() was called on tracefs that was previously mounted, and was not closed. It will leave a "cursor dentry" in the subdirs list of the current dentries that set_gid() walks. On a remount of tracefs, the container_of(dentry->d_inode) will dereference a NULL pointer and cause a crash when referenced.

Simply have a check for dentry->d_inode to see if it is NULL and if so, skip that entry.

- Fix the bits of the eventfs_inode structure. The "is_events" bit was taken from the nr_entries field, but the nr_entries field wasn't updated to be 30 bits and was still 31. Including the "is_freed" bit this would use 33 bits which would make the structure use another integer for just one bit.

Steven Rostedt (Google) (2):
tracefs: Check for dentry->d_inode exists in set_gid()
eventfs: Fix bitwise fields for "is_events"

----

 fs/tracefs/inode.c    | 4 ++++
 fs/tracefs/internal.h | 2 +-
 2 files changed, 5 insertions(+), 1 deletion(-)
author maintainer
#1 lore raw

PATCH 1/2 tracefs: Check for dentry->d_inode exists in set_gid()

From: "Steven Rostedt (Google)" <[email protected]>

If a getdents() is called on the tracefs directory but does not get all the files, it can leave a "cursor" dentry in the d_subdirs list of tracefs dentry. This cursor dentry does not have a d_inode for it. Before referencing tracefs_inode from the dentry, the d_inode must first be checked if it has content. If not, then it's not a tracefs_inode and can be ignored.

The following caused a crash:

#define getdents64(fd, dirp, count) syscall(SYS_getdents64, fd, dirp, count) #define BUF_SIZE 256
#define TDIR "/tmp/file0"

int main(void)

 {
	char buf[BUF_SIZE];
	int fd;
       	int n;

       	mkdir(TDIR, 0777);
	mount(NULL, TDIR, "tracefs", 0, NULL);
       	fd = openat(AT_FDCWD, TDIR, O_RDONLY);
       	n = getdents64(fd, buf, BUF_SIZE);
       	ret = mount(NULL, TDIR, NULL, MS_NOSUID|MS_REMOUNT|MS_RELATIME|MS_LAZYTIME,
		    "gid=1000");
	return 0;

}

That's because the 256 BUF_SIZE was not big enough to read all the dentries of the tracefs file system and it left a "cursor" dentry in the subdirs of the tracefs root inode. Then on remounting with "gid=1000", it would cause an iteration of all dentries which hit:

	ti = get_tracefs(dentry->d_inode);
	if (ti && (ti->flags & TRACEFS_EVENT_INODE))
		eventfs_update_gid(dentry, gid);

Which crashed because of the dereference of the cursor dentry which had a NULL d_inode.

In the subdir loop of the dentry lookup of set_gid(), if a child has a NULL d_inode, simply skip it.

Cc Masami Hiramatsu <[email protected]>
Cc Mark Rutland <[email protected]>
Cc Mathieu Desnoyers <[email protected]>
Fixes 7e8358edf503e ("eventfs: Fix file and directory uid and gid ownership")
---
 fs/tracefs/inode.c | 4 ++++
 1 file changed, 4 insertions(+)
fs/tracefs/inode.c +4 -0
index 62524b20964e..bc86ffdb103b 100644
@@ -215,6 +215,10 @@ static void set_gid(struct dentry *parent, kgid_t gid)
 		struct dentry *dentry = list_entry(tmp, struct dentry, d_child);
 		next = tmp->next;
 
+		/* Note, getdents() can add a cursor dentry with no inode */
+		if (!dentry->d_inode)
+			continue;
+
 		spin_lock_nested(&dentry->d_lock, DENTRY_D_LOCK_NESTED);
 
 		change_gid(dentry, gid);
Signature · 1 line
2.42.0
author maintainer
#2 lore raw

PATCH 2/2 eventfs: Fix bitwise fields for "is_events"

In reply to Steven Rostedt

From: "Steven Rostedt (Google)" <[email protected]>

A flag was needed to denote which eventfs_inode was the "events" directory, so a bit was taken from the "nr_entries" field, as there's not that many entries, and 2^30 is plenty. But the bit number for nr_entries was not updated to reflect the bit taken from it, which would add an unnecessary integer to the structure.

Cc Masami Hiramatsu <[email protected]>
Cc Mathieu Desnoyers <[email protected]>
Fixes 7e8358edf503e ("eventfs: Fix file and directory uid and gid ownership")
---
 fs/tracefs/internal.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
fs/tracefs/internal.h +1 -1
index 899e447778ac..42bdeb471a07 100644
@@ -63,7 +63,7 @@ struct eventfs_inode {
 	};
 	unsigned int			is_freed:1;
 	unsigned int			is_events:1;
-	unsigned int			nr_entries:31;
+	unsigned int			nr_entries:30;
 };
 
 static inline struct tracefs_inode *get_tracefs(const struct inode *inode)
Signature · 1 line
2.42.0