Skip to contentSkip to conversation

tracing: Fix memory leak in tracing_err_log_open()

PATCH In review Takeshi Misawa 2 participants 2 messages Jun 28, 2019 Tracing

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 tracing: Fix memory leak in tracing_err_log_open()

If tracing_err_log_open() call seq_open(), allocated memory is not freed.

kmemleak report:

unreferenced object 0xffff92c0781d1100 (size 128):
comm "tail", pid 15116, jiffies 4295163855 (age 22.704s) hex dump (first 32 bytes):

    00 f0 08 e5 c0 92 ff ff 00 10 00 00 00 00 00 00  ................
    00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
  backtrace:
    [<000000000d0687d5>] kmem_cache_alloc+0x11f/0x1e0
    [<000000003e3039a8>] seq_open+0x2f/0x90
    [<000000008dd36b7d>] tracing_err_log_open+0x67/0x140
    [<000000005a431ae2>] do_dentry_open+0x1df/0x3a0
    [<00000000a2910603>] vfs_open+0x2f/0x40
    [<0000000038b0a383>] path_openat+0x2e8/0x1690
    [<00000000fe025bda>] do_filp_open+0x9b/0x110
    [<00000000483a5091>] do_sys_open+0x1ba/0x260
    [<00000000c558b5fd>] __x64_sys_openat+0x20/0x30
    [<000000006881ec07>] do_syscall_64+0x5a/0x130
    [<00000000571c2e94>] entry_SYSCALL_64_after_hwframe+0x44/0xa9

Fix this by calling seq_release() in tracing_err_log_fops.release().

---
Dear Steven Rostedt

Thanks for reviewing.

Quoted 7 lines

Actually, I think it is safer to have the condition be:

        if (file->f_mode & FMODE_READ)

As that would match the open.

Can you send a v2?

I send a v2 patch.

Regards.

---
 kernel/trace/trace.c | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)
kernel/trace/trace.c +13 -1
index 83e08b78dbee..4122ccde6ec2 100644
@@ -7126,12 +7126,24 @@ static ssize_t tracing_err_log_write(struct file *file,
 	return count;
 }
 
+static int tracing_err_log_release(struct inode *inode, struct file *file)
+{
+	struct trace_array *tr = inode->i_private;
+
+	trace_array_put(tr);
+
+	if (file->f_mode & FMODE_READ)
+		seq_release(inode, file);
+
+	return 0;
+}
+
 static const struct file_operations tracing_err_log_fops = {
 	.open           = tracing_err_log_open,
 	.write		= tracing_err_log_write,
 	.read           = seq_read,
 	.llseek         = seq_lseek,
-	.release	= tracing_release_generic_tr,
+	.release        = tracing_err_log_release,
 };
 
 static int tracing_buffers_open(struct inode *inode, struct file *filp)
Signature · 1 line
2.17.1
Reviewed-by on v2 1/1
#1 lore raw

Hi,

Quoted 67 lines · Takeshi Misawa

If tracing_err_log_open() call seq_open(), allocated memory is not freed.

kmemleak report:

unreferenced object 0xffff92c0781d1100 (size 128):
comm "tail", pid 15116, jiffies 4295163855 (age 22.704s) hex dump (first 32 bytes):

    00 f0 08 e5 c0 92 ff ff 00 10 00 00 00 00 00 00  ................
    00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
  backtrace:
    [<000000000d0687d5>] kmem_cache_alloc+0x11f/0x1e0
    [<000000003e3039a8>] seq_open+0x2f/0x90
    [<000000008dd36b7d>] tracing_err_log_open+0x67/0x140
    [<000000005a431ae2>] do_dentry_open+0x1df/0x3a0
    [<00000000a2910603>] vfs_open+0x2f/0x40
    [<0000000038b0a383>] path_openat+0x2e8/0x1690
    [<00000000fe025bda>] do_filp_open+0x9b/0x110
    [<00000000483a5091>] do_sys_open+0x1ba/0x260
    [<00000000c558b5fd>] __x64_sys_openat+0x20/0x30
    [<000000006881ec07>] do_syscall_64+0x5a/0x130
    [<00000000571c2e94>] entry_SYSCALL_64_after_hwframe+0x44/0xa9

Fix this by calling seq_release() in tracing_err_log_fops.release().

---
Dear Steven Rostedt

Thanks for reviewing.

Quoted 7 lines

Actually, I think it is safer to have the condition be:

        if (file->f_mode & FMODE_READ)

As that would match the open.

Can you send a v2?

I send a v2 patch.

Regards.

---
 kernel/trace/trace.c | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)
kernel/trace/trace.c +10 -0
index 83e08b78dbee..4122ccde6ec2 100644
@@ -7126,12 +7126,24 @@ static ssize_t tracing_err_log_write(struct
 file *file,
 	return count;
 }
 
+static int tracing_err_log_release(struct inode *inode, struct file
 *file)
+{
+	struct trace_array *tr = inode->i_private;
+
+	trace_array_put(tr);
+
+	if (file->f_mode & FMODE_READ)
+		seq_release(inode, file);
+
+	return 0;

kernel/trace/trace.c static ssize_t tracing_err_log_write(struct · line 7149

Quoted 10 lines
kernel/trace/trace.c +3 -1
+}
+
 static const struct file_operations tracing_err_log_fops = {
 	.open           = tracing_err_log_open,
 	.write		= tracing_err_log_write,
 	.read           = seq_read,
 	.llseek         = seq_lseek,
-	.release	= tracing_release_generic_tr,
+	.release        = tracing_err_log_release,
 };

v2 looks good to me, thanks for sending this fix.

Quoted 2 lines

static int tracing_buffers_open(struct inode *inode, struct file
*filp)