Re: [PATCH v5 5/8] fat: restructure export_operations

[Namjae Jeon]

2012/12/4, OGAWA Hirofumi <hirofumi@xxxxxxxxxxxxxxxxxx>:
> Namjae Jeon <linkinjeon@xxxxxxxxx> writes:
>
>>>> +static struct dentry *fat_fh_to_dentry_nostale(struct super_block *sb,
>>>> +					       struct fid *fh, int fh_len,
>>>> +					       int fh_type)
>>>> +{
>>>> +	struct inode *inode = NULL;
>>>> +	struct fat_fid *fid = (struct fat_fid *)fh;
>>>> +	loff_t i_pos;
>>>> +
>>>> +	switch (fh_type) {
>>>> +	case FILEID_FAT_WITHOUT_PARENT:
>>>> +		if (fh_len < FAT_FID_SIZE_WITHOUT_PARENT)
>>>> +			return NULL;
>>>> +	case FILEID_FAT_WITH_PARENT:
>>>> +		if ((fh_len < FAT_FID_SIZE_WITH_PARENT) &&
>>>> +			(fh_type == FILEID_FAT_WITH_PARENT))
>>>> +			return NULL;
>>>
>>> Do we have to care (FILEID_FAT_WITH_PARENT and fh_len < 5) here?
>>>
>>> 	if (fh_len < 2)
>>> 		return NULL;
>>>
>>> 	switch (fh_type) {
>>> 	case FILEID_INO32_GEN:
>>> 	case FILEID_INO32_GEN_PARENT:
>>> 		inode = get_inode(sb, fid->i32.ino, fid->i32.gen);
>>> 		break;
>>> 	}
>>>
>>> 	return d_obtain_alias(inode);
>>>
>>> generic_fh_to_dentry() is above. I wonder why we have to care
>>> fat_fid->parent* here.
>> Let me think, if âsubtreeâ checking is enabled then we should check
>> the length condition over here also? Please share if there are any
>> other comments also.
>
> I'm not sure what did you mean. Where is "subtree" check you are
> talking? This is fh_to_dentry(), so we don't use parent at all, so
> length == 3 is enough?
With fileID type="FILEID_FAT_WITHOUT_PARENT", fhlen  should be '3'
With fileId type="FILEID_FAT_WITH_PARENT", fhlen should be '5'

While encoding, WITH_PARENT is selected when subtree check is enabled
on the NFS Server.

So, when decoding request is arrived-  fileid type will be among the '2' cases:
Now, in case of fh_to_dentry() - when we consider, that the reqquest
for fileid type WITH_PARENT
then i think the conditions in fh_to_dentry should be:

if((fh_type == FILEID_FAT_WITH_PARENT) && fh_len != 5)
	return NULL;
else if (fh_len != 3)
	return NULL;

So, we took care of these '2' conditions within the switch statement
based on the 'fh_type'. We can just change the comparision condition
from '<' to '!=':
switch (fh_type) {
+	case FILEID_FAT_WITHOUT_PARENT:
+		if (fh_len != FAT_FID_SIZE_WITHOUT_PARENT)
+			return NULL;
+	case FILEID_FAT_WITH_PARENT:
+		if ((fh_len != FAT_FID_SIZE_WITH_PARENT) &&
+			(fh_type == FILEID_FAT_WITH_PARENT))
+			return NULL;
+		i_pos = fid->i_pos_hi;
+		i_pos = (i_pos << 32) | (fid->i_pos_low);
+		inode = __fat_nfs_get_inode(sb, 0, fid->i_gen, i_pos);
+		break;
+	}

I think there is no need to push the comparision statements in the
begining similar to generic_fh_to_dentry.

Thanks OGAWA.
> --
> OGAWA Hirofumi <hirofumi@xxxxxxxxxxxxxxxxxx>
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/