Messages in this thread | | | Date | Sun, 13 Jul 1997 14:31:53 +0200 | From | "Volker.Lendecke" <> | Subject | dcache implications for smbfs? |
| |
-----BEGIN PGP SIGNED MESSAGE-----
Hello!
As the VFS is currently massively rewritten, I have a question regarding smbfs. In smbfs you only have the file path. It is great that the complete path to a file can now be accessed with the dcache.
But for smbfs, iget is still a problem, as smbfs has to invent i-node numbers. To do this, I would like to randomly pick one, scan all the i-nodes specific to a mount point, and pick another one if I find the one I chose. Do I have to scan all the i-nodes? (I do not like the 2^-32 chance of a i-node number clash..)
As far as I can see, read_inode is only called via the fs-specific lookup routine. Is this correct? If it is, I would *love* to pass a pointer from smb_lookup to smb_read_inode. I think nfs would benefit from this as well, and ncpfs would definitely. As all file systems have to be touched anyway, is there any chance for this change?
Volker
-----BEGIN PGP SIGNATURE----- Version: 2.6.3i Charset: noconv
iQCVAwUBM8jKdj/9BWnmOc5FAQFqbgP/UZkO36ro2NmWcEcJAbqCmekC1ms6L/Rn Xm9/L9fU65BTB4oMcU/Qrr8jPLf0QRlKld5UpL7/35MUJNBS9fw6rb9XrJ7sDrDo I2D1WNUwZnttLhhJIspKtfNzUI0jO/U/HhxjIUDACpCsjx6qX19mDmXO5ATteAr/ r0rzi6LelG0= =cXN6 -----END PGP SIGNATURE-----
| |