Re: [CFT][PATCH] kernfs: Correct kernfs directory seeks.

[Eric W. Biederman]

ebiederm@xxxxxxxxxxxx (Eric W. Biederman) writes:

> "Hatayama, Daisuke" <d.hatayama@xxxxxxxxxxxxxx> writes:
>
>>> Can you test this and please verify it fixes your issue?
>>
>> I tried this patch on top of v4.17 but the system fails to boot
>> without detecting root disks by dracut like this:
[snip]

>> OTOH, there's no issue on the pure v4.17 kernel.
>>
>> As above, ls /sys/module looks apparently good. But I guess any part of
>> behavior of getdentries() on sysfs must have changed, affecting the disk
>> detection...
>
> I haven't been able to reproduce this yet.  My test system boots fine.
> Which fedora are you testing on?

I reproduced something similar and fedora 28.  So I think I have found
and fixed the issue.  I believe I simply reversed the test at the end of
kernfs_dir_pos. AKA "<" instead of ">".

I am going to see if I can test my changes more throughly on this side
and then repost.



>>>  fs/kernfs/dir.c | 109
>>> ++++++++++++++++++++++++++++++++++----------------------
>>>  1 file changed, 67 insertions(+), 42 deletions(-)
>>> 
>>> diff --git a/fs/kernfs/dir.c b/fs/kernfs/dir.c
>>> index 89d1dc19340b..8148b5fec48d 100644
>>> --- a/fs/kernfs/dir.c
>>> +++ b/fs/kernfs/dir.c
>>> @@ -1584,53 +1584,75 @@ static int kernfs_dir_fop_release(struct inode *inode,
>>> struct file *filp)
>>>  	return 0;
>>>  }
>>> 
>>> +static struct kernfs_node *kernfs_dir_next(struct kernfs_node *pos)
>>> +{
>>> +	struct rb_node *node = rb_next(&pos->rb);
>>> +	return node ? rb_to_kn(node) : NULL;
>>> +}
>>> +
>>>  static struct kernfs_node *kernfs_dir_pos(const void *ns,
>>> -	struct kernfs_node *parent, loff_t hash, struct kernfs_node *pos)
>>> +	struct kernfs_node *parent, loff_t off, struct kernfs_node *saved)
>>>  {
>>> -	if (pos) {
>>> -		int valid = kernfs_active(pos) &&
>>> -			pos->parent == parent && hash == pos->hash;
>>> -		kernfs_put(pos);
>>> -		if (!valid)
>>> -			pos = NULL;
>>> -	}
>>> -	if (!pos && (hash > 1) && (hash < INT_MAX)) {
>>> -		struct rb_node *node = parent->dir.children.rb_node;
>>> -		while (node) {
>>> -			pos = rb_to_kn(node);
>>> -
>>> -			if (hash < pos->hash)
>>> -				node = node->rb_left;
>>> -			else if (hash > pos->hash)
>>> -				node = node->rb_right;
>>> -			else
>>> -				break;
>>> +	struct kernfs_node *pos;
>>> +	struct rb_node *node;
>>> +	unsigned int hash;
>>> +	const char *name = "";
>>> +
>>> +	/* Is off a valid name hash? */
>>> +	if ((off < 2) || (off >= INT_MAX))
>>> +		return NULL;
>>> +	hash = off;
>>> +
>>> +	/* Is the saved position usable? */
>>> +	if (saved) {
>>> +		/* Proper parent and hash? */
>>> +		if ((parent != saved->parent) || (saved->hash != hash)) {
>>> +			saved = NULL;
>>
>> name is uninitialized in this path.
>
> It is.  name is initialized to "" see above.
>
>>> +		} else {
>>> +			if (kernfs_active(saved))
>>> +				return saved;
>>> +			name = saved->name;
>>>  		}
>>>  	}
>>> -	/* Skip over entries which are dying/dead or in the wrong namespace
>>> */
>>> -	while (pos && (!kernfs_active(pos) || pos->ns != ns)) {
>>> -		struct rb_node *node = rb_next(&pos->rb);
>>> -		if (!node)
>>> -			pos = NULL;
>>> +
>>> +	/* Find the closest pos to the hash we are looking for */
>>> +	pos = NULL;
>>> +	node = parent->dir.children.rb_node;
>>> +	while (node) {
>>> +		int result;
>>> +
>>> +		pos = rb_to_kn(node);
>>> +		result = kernfs_name_compare(hash, name, ns, pos);
>>> +		if (result < 0)
>>> +			node = node->rb_left;
>>> +		else if (result > 0)
>>> +			node = node->rb_right;
>>>  		else
>>> -			pos = rb_to_kn(node);
>>> +			break;
>>>  	}
>>> +
>>> +	/* Ensure pos is at or beyond the target position */
>>> +	if (pos && (kernfs_name_compare(hash, name, ns, pos) < 0))
                                                    ^^^^^^^^^^^^^^^^
                                          should be > 0
>>> +		pos = kernfs_dir_next(pos);
>>> +

Eric