Background:
I have an Amazon S3 bucket with data uploaded using the AWS web interface. Now I need to give permissions-based access to paths within my buckets to FTP users. To do this, I'm running vsftpd and s3fs on a fresh Amazon EC2 instance. Problem is that bucket contents have the pathname instead of a dotfile and the huge filesize creates an exception in vsftpd. ls -la
looks like this:
$ pwd
/mnt/my-bucket-name/pathname
$ ls -la
total 1
drwxrwxr-x 1 root root 0 Jun 16 19:54 dir1
drwxrwxr-x 1 root root 0 Jun 16 19:54 dir2
---------- 1 root root 18446744073709551615 Dec 31 1969 pathname
Whereas it should look more like:
$ ls -la
total 500
drwxrwxr-x 1 root root 500 Jun 16 19:54 .
drwxrwxr-x 1 root root 500 Jun 16 19:54 ..
drwxrwxr-x 1 root root 500 Jun 16 19:54 dir1
drwxrwxr-x 1 root root 500 Jun 16 19:54 dir2
When FTP users try to connect they get an error:
ftp> ls
229 Entering Extended Passive Mode (|||14017|).
150 Here comes the directory listing.
500 OOPS: invalid inode size in vsf_sysutil_statbuf_get_size
The bucket is mounted with this command: sudo s3fs my-bucket-name /mnt/my-bucket-name -o allow_other -o default_acl=public-read
I created a ticket on the s3fs google code page a month ago, hacked at this for hours and I'm stumped. Any ideas what I'm doing wrong?