[RFC PATCH 1/2] tracing: fixes of ftrace_enable_fops

[Gabriele Paoloni]

Currently there are different issues associated with ftrace_enable_fops
- event_enable_write: *ppos is increased while not used at all in the
  write operation itself (following a write, this could lead a read to
  fail or report a corrupted event status);
- event_enable_read: cnt < strlen(buf) is allowed and this can lead to
  reading an incomplete event status (i.e. not all status characters
  are retrieved) and/or reading the status in a non-atomic way (i.e.
  the status could change between two consecutive reads);
- .llseek is set to default_llseek: this is wrong since for this
  type of files it does not make sense to reposition the ppos offset.
  Hence this should be set instead to noop_llseek.

This patch fixes all the issues listed above.

Signed-off-by: Gabriele Paoloni <gpaoloni@xxxxxxxxxx>
Tested-by: Alessandro Carminati <acarmina@xxxxxxxxxx>
---
 kernel/trace/trace_events.c | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c
index 120531268abf..5e84ef01d0c8 100644
--- a/kernel/trace/trace_events.c
+++ b/kernel/trace/trace_events.c
@@ -1798,6 +1798,13 @@ event_enable_read(struct file *filp, char __user *ubuf, size_t cnt,
 
 	strcat(buf, "\n");
 
+	/*
+	 * A requested cnt less than strlen(buf) could lead to a wrong
+	 * event status being reported.
+	 */
+	if (cnt < strlen(buf))
+		return -EINVAL;
+
 	return simple_read_from_buffer(ubuf, cnt, ppos, buf, strlen(buf));
 }
 
@@ -1833,8 +1840,6 @@ event_enable_write(struct file *filp, const char __user *ubuf, size_t cnt,
 		return -EINVAL;
 	}
 
-	*ppos += cnt;
-
 	return cnt;
 }
 
@@ -2557,7 +2562,7 @@ static const struct file_operations ftrace_enable_fops = {
 	.read = event_enable_read,
 	.write = event_enable_write,
 	.release = tracing_release_file_tr,
-	.llseek = default_llseek,
+	.llseek = noop_llseek,
 };
 
 static const struct file_operations ftrace_event_format_fops = {
-- 
2.48.1