2007-08-02 06:42:01

by Satyam Sharma

[permalink] [raw]
Subject: Re: [PATCH] single_open/seq_release leak diagnostics

Hi,


On Tue, 31 Jul 2007, Alexey Dobriyan wrote:

> [...]
> --- a/fs/seq_file.c
> +++ b/fs/seq_file.c
> @@ -281,6 +281,13 @@ EXPORT_SYMBOL(seq_lseek);
> int seq_release(struct inode *inode, struct file *file)
> {
> struct seq_file *m = (struct seq_file *)file->private_data;
> +
> + if (m->seq_ops_allocated) {
> + struct dentry *dentry = file->f_dentry;
> + printk("memory leak: '%.*s'\n",
> + dentry->d_name.len, dentry->d_name.name);
> + WARN_ON(1);
> + }
> kfree(m->buf);
> kfree(m);
> return 0;
> @@ -401,9 +408,12 @@ int single_open(struct file *file, int (*show)(struct seq_file *, void *),
> op->stop = single_stop;
> op->show = show;
> res = seq_open(file, op);
> - if (!res)
> - ((struct seq_file *)file->private_data)->private = data;
> - else
> + if (!res) {
> + struct seq_file *seq = file->private_data;
> +
> + seq->private = data;
> + seq->seq_ops_allocated = 1;
> + } else
> kfree(op);
> }
> return res;
> @@ -412,8 +422,13 @@ EXPORT_SYMBOL(single_open);
>
> int single_release(struct inode *inode, struct file *file)
> {
> - const struct seq_operations *op = ((struct seq_file *)file->private_data)->op;
> - int res = seq_release(inode, file);
> + struct seq_file *seq = file->private_data;
> + const struct seq_operations *op = seq->op;
> + int res;
> +
> + /* All roads lead to seq_release(), so... */
> + seq->seq_ops_allocated = 0;
> + res = seq_release(inode, file);
> kfree(op);
> return res;
> }
> --- a/include/linux/seq_file.h
> +++ b/include/linux/seq_file.h
> @@ -22,6 +22,7 @@ struct seq_file {
> struct mutex lock;
> const struct seq_operations *op;
> void *private;
> + unsigned int seq_ops_allocated:1;
> };
>
> struct seq_operations {

Hmm, curiously, I think this patch just killed the utility of having
single_release() around in the first place :-)

We might as well free the seq_file->op when we detect that we're
leaking it in seq_release() itself. That makes single_release() wholly
redundant to keep, and we can just convert all its users to seq_release()
itself. With less of these around, lesser probability of someone coding
a bug/leak in the first place!


Just my Rs. 0.02,

Satyam