bpf: Do cleanup in bpf_bprintf_cleanup only when needed

Currently we always cleanup/decrement bpf_bprintf_nest_level variable
in bpf_bprintf_cleanup if it's > 0.

There's possible scenario where this could cause a problem, when
bpf_bprintf_prepare does not get bin_args buffer (because num_args is 0)
and following bpf_bprintf_cleanup call decrements bpf_bprintf_nest_level
variable, like:

  in task context:
    bpf_bprintf_prepare(num_args != 0) increments 'bpf_bprintf_nest_level = 1'
    -> first irq :
       bpf_bprintf_prepare(num_args == 0)
       bpf_bprintf_cleanup decrements 'bpf_bprintf_nest_level = 0'
    -> second irq:
       bpf_bprintf_prepare(num_args != 0) bpf_bprintf_nest_level = 1
       gets same buffer as task context above

Adding check to bpf_bprintf_cleanup and doing the real cleanup only if we
got bin_args data in the first place.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Yonghong Song <yhs@fb.com>
Link: https://lore.kernel.org/bpf/20221215214430.1336195-3-jolsa@kernel.org

3 files changed, 13 insertions, 11 deletions

diff --git a/include/linux/bpf.h b/include/linux/bpf.h
index cc390ba32e70..656879385fbf 100644
--- a/include/linux/bpf.h
+++ b/include/linux/bpf.h
@@ -2803,7 +2803,7 @@ struct bpf_bprintf_data {
 
 int bpf_bprintf_prepare(char *fmt, u32 fmt_size, const u64 *raw_args,
 			u32 num_args, struct bpf_bprintf_data *data);
-void bpf_bprintf_cleanup(void);
+void bpf_bprintf_cleanup(struct bpf_bprintf_data *data);
 
 /* the implementation of the opaque uapi struct bpf_dynptr */
 struct bpf_dynptr_kern {
diff --git a/kernel/bpf/helpers.c b/kernel/bpf/helpers.c
index 7dbf6bb72cad..9cca02e13f2e 100644
--- a/kernel/bpf/helpers.c
+++ b/kernel/bpf/helpers.c
@@ -784,12 +784,14 @@ static int try_get_fmt_tmp_buf(char **tmp_buf)
 	return 0;
 }
 
-void bpf_bprintf_cleanup(void)
+void bpf_bprintf_cleanup(struct bpf_bprintf_data *data)
 {
-	if (this_cpu_read(bpf_bprintf_nest_level)) {
-		this_cpu_dec(bpf_bprintf_nest_level);
-		preempt_enable();
-	}
+	if (!data->bin_args)
+		return;
+	if (WARN_ON_ONCE(this_cpu_read(bpf_bprintf_nest_level) == 0))
+		return;
+	this_cpu_dec(bpf_bprintf_nest_level);
+	preempt_enable();
 }
 
 /*
@@ -1021,7 +1023,7 @@ nocopy_fmt:
 	err = 0;
 out:
 	if (err)
-		bpf_bprintf_cleanup();
+		bpf_bprintf_cleanup(data);
 	return err;
 }
 
@@ -1047,7 +1049,7 @@ BPF_CALL_5(bpf_snprintf, char *, str, u32, str_size, char *, fmt,
 
 	err = bstr_printf(str, str_size, fmt, data.bin_args);
 
-	bpf_bprintf_cleanup();
+	bpf_bprintf_cleanup(&data);
 
 	return err + 1;
 }
diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
index 3e849c3a7cc8..2129f7c68bb5 100644
--- a/kernel/trace/bpf_trace.c
+++ b/kernel/trace/bpf_trace.c
@@ -396,7 +396,7 @@ BPF_CALL_5(bpf_trace_printk, char *, fmt, u32, fmt_size, u64, arg1,
 	trace_bpf_trace_printk(buf);
 	raw_spin_unlock_irqrestore(&trace_printk_lock, flags);
 
-	bpf_bprintf_cleanup();
+	bpf_bprintf_cleanup(&data);
 
 	return ret;
 }
@@ -454,7 +454,7 @@ BPF_CALL_4(bpf_trace_vprintk, char *, fmt, u32, fmt_size, const void *, args,
 	trace_bpf_trace_printk(buf);
 	raw_spin_unlock_irqrestore(&trace_printk_lock, flags);
 
-	bpf_bprintf_cleanup();
+	bpf_bprintf_cleanup(&data);
 
 	return ret;
 }
@@ -494,7 +494,7 @@ BPF_CALL_5(bpf_seq_printf, struct seq_file *, m, char *, fmt, u32, fmt_size,
 
 	seq_bprintf(m, fmt, data.bin_args);
 
-	bpf_bprintf_cleanup();
+	bpf_bprintf_cleanup(&data);
 
 	return seq_has_overflowed(m) ? -EOVERFLOW : 0;
 }