Re: [PATCH] eventfs: Have inodes have unique inode numbers

From: Steven Rostedt <rostedt@goodmis.org>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>,
	Mathieu Desnoyers <mathieu.desnoyers@efficios.com>,
	LKML <linux-kernel@vger.kernel.org>,
	Linux Trace Devel <linux-trace-devel@vger.kernel.org>,
	Christian Brauner <brauner@kernel.org>,
	Ajay Kaher <ajay.kaher@broadcom.com>,
	Geert Uytterhoeven <geert@linux-m68k.org>,
	linux-fsdevel <linux-fsdevel@vger.kernel.org>
Subject: Re: [PATCH] eventfs: Have inodes have unique inode numbers
Date: Sun, 28 Jan 2024 22:40:54 -0500	[thread overview]
Message-ID: <20240128224054.0df489b8@rorschach.local.home> (raw)
In-Reply-To: <20240128213249.605a7ade@rorschach.local.home>

On Sun, 28 Jan 2024 21:32:49 -0500
Steven Rostedt <rostedt@goodmis.org> wrote:

>  # echo 'p:sched schedule' >> kprobe_events
>  # ls events/kprobes
> enable  filter  sched  timer
> 
>  # ls events/kprobes/sched/
> ls: reading directory 'events/kprobes/sched/': Invalid argument
> 
> I have no access to the directory that was deleted and recreated.

Ah, this was because the final iput() does dentry->d_fsdata = NULL, and
in the lookup code I have:

	mutex_lock(&eventfs_mutex);
	ei = READ_ONCE(ti->private);
	if (ei && ei->is_freed)
		ei = NULL;
	mutex_unlock(&eventfs_mutex);

	if (!ei) {
		printk("HELLO no ei\n");
		goto out;
	}

Where that printk() was triggering.

So at least it's not calling back into the tracing code ;-)

Interesting that it still did the lookup, even though it was already
referenced.

I'm still learning the internals of VFS.

Anyway, after keeping the d_fsdata untouched (not going to NULL), just
to see what would happen, I ran it again with KASAN and did trigger:

[  106.255468] ==================================================================
[  106.258400] BUG: KASAN: slab-use-after-free in tracing_open_file_tr+0x3a/0x120
[  106.261228] Read of size 8 at addr ffff8881136f27b8 by task cat/868

[  106.264506] CPU: 2 PID: 868 Comm: cat Not tainted 6.8.0-rc1-test-00008-gbee668990ac4-dirty #454
[  106.267810] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
[  106.271337] Call Trace:
[  106.272406]  <TASK>
[  106.273317]  dump_stack_lvl+0x5c/0xc0
[  106.274750]  print_report+0xcf/0x670
[  106.276173]  ? __virt_addr_valid+0x15a/0x330
[  106.278807]  kasan_report+0xd8/0x110
[  106.280172]  ? tracing_open_file_tr+0x3a/0x120
[  106.281745]  ? tracing_open_file_tr+0x3a/0x120
[  106.283343]  tracing_open_file_tr+0x3a/0x120
[  106.284887]  do_dentry_open+0x3b7/0x950
[  106.286284]  ? __pfx_tracing_open_file_tr+0x10/0x10
[  106.287992]  path_openat+0xea8/0x11d0

That was with just these commands:

  cd /sys/kernel/tracing/
  echo 'p:sched schedule' >> /sys/kernel/tracing/kprobe_events 
  echo 'p:timer read_current_timer' >> kprobe_events 
  ls events/kprobes/
  cat events/kprobes/sched/enable
  ls events/kprobes/sched
  echo '-:sched schedule' >> /sys/kernel/tracing/kprobe_events 
  ls events/kprobes/sched/enable
  cat events/kprobes/sched/enable

BTW, the ls after the deletion returned:

 # ls events/kprobes/sched/enable
 events/kprobes/sched/enable

In a normal file system that would be equivalent to:

 # mkdir events/kprobes/sched
 # touch events/kprobes/sched/enable
 # rm -rf events/kprobes/sched
 # ls events/kprobes/sched/enable
 events/kprobes/sched/enable

-- Steve