Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Subject: Re: [PATCH v2 6/9] debugfs: Implement debugfs_create_str()
To:     Peter Zijlstra <peterz@infradead.org>, Greg KH <greg@kroah.com>
Cc:     mingo@kernel.org, mgorman@suse.de, juri.lelli@redhat.com,
        vincent.guittot@linaro.org, dietmar.eggemann@arm.com,
        rostedt@goodmis.org, bsegall@google.com, bristot@redhat.com,
        joshdon@google.com, valentin.schneider@arm.com,
        linux-kernel@vger.kernel.org
References: <20210326103352.603456266@infradead.org>
 <20210326103935.183934395@infradead.org> <YF2/41K4xs3ZOQdV@kroah.com>
 <YF3DF+T8nPRgt7Ao@hirez.programming.kicks-ass.net>
 <YF3F0JbbEpeSGzW6@kroah.com>
 <YF3Hv5zXb/6lauzs@hirez.programming.kicks-ass.net>
From:   Rasmus Villemoes <linux@rasmusvillemoes.dk>
Message-ID: <70594935-18e6-0791-52f9-0448adf37155@rasmusvillemoes.dk>
Date:   Fri, 26 Mar 2021 13:53:59 +0100
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101
 Thunderbird/78.7.1
MIME-Version: 1.0
In-Reply-To: <YF3Hv5zXb/6lauzs@hirez.programming.kicks-ass.net>
Content-Type: text/plain; charset=windows-1252
Content-Language: en-US
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 26/03/2021 12.38, Peter Zijlstra wrote:

> Implement debugfs_create_str() to easily display names and such in
> debugfs.

Patches 7-9 don't seem to add any users of this. What's it for precisely?

> +
> +again:
> +	rcu_read_lock();
> +	str = rcu_dereference(*(char **)file->private_data);
> +	len = strlen(str) + 1;
> +
> +	if (!copy || copy_len < len) {
> +		rcu_read_unlock();
> +		kfree(copy);
> +		copy = kmalloc(len + 1, GFP_KERNEL);
> +		if (!copy) {
> +			debugfs_file_put(dentry);
> +			return -ENOMEM;
> +		}
> +		copy_len = len;
> +		goto again;
> +	}
> +
> +	strncpy(copy, str, len);
> +	copy[len] = '\n';
> +	copy[len+1] = '\0';
> +	rcu_read_unlock();

As noted (accidentally off-list), this is broken. I think you want this
on top

- len = strlen(str) + 1;
+ len = strlen(str);

- strncpy(copy, str, len);
+ memcpy(copy, str, len);
  copy[len] = '\n';
- copy[len+1] = '\0';

> +EXPORT_SYMBOL_GPL(debugfs_read_file_str);

Why?

> +
> +ssize_t debugfs_write_file_str(struct file *file, const char __user *user_buf,
> +			       size_t count, loff_t *ppos)
> +{
> +	struct dentry *dentry = F_DENTRY(file);
> +	char *old, *new = NULL;
> +	int pos = *ppos;
> +	int r;
> +
> +	r = debugfs_file_get(dentry);
> +	if (unlikely(r))
> +		return r;
> +
> +	old = *(char **)file->private_data;
> +
> +	/* only allow strict concattenation */
> +	r = -EINVAL;
> +	if (pos && pos != strlen(old))
> +		goto error;

Other than the synchronize_rcu() below, I don't see any rcu stuff in
here. What prevents one task from kfree'ing old while another computes
its strlen()? Or two tasks from getting the same value of old and both
kfree'ing the same pointer?

And what part of the kernel periodically looks at some string and
decides something needs to be done? Shouldn't a write imply some
notification or callback? I can see the usefulness of the read part to
expose some otherwise-maintained string, but what good does allowing
writes do?

Rasmus