View source

From Ext4

You do not have permission to edit this page, for the following reasons:

• The action you have requested is limited to users in the group: Users.
• You must confirm your e-mail address before editing pages. Please set and validate your e-mail address through your user preferences.

You can view and copy the source of this page:

== New Features for discussion: == Not sure if this is the place, but for your discussion here is my proposal for a feature: SAN Bacup aware FS My thinking goes like thus; When in a SAN environment a backup server still has to ask the host with data to be backed up to send it the data. If the FS were multiple access aware then it could freeze the data and record to journal only any changes untill the backup is done. I see this happening as follows: #. A part of the FS structure is reserved for another host to write a predefined patern. #. If the patern is written to that area then: ##. The file system will guarantee that after a set number of miliseconds all new writes will be to journal only. ##. The FS will update another reserved structure to reflect the journaling status. #. A different patern written to the same structure will ask for the FS to return to normal and flush the journal to normal disk locations. Thus a remote backup host could; ask for the FS to start journal only operation, find out it has done so, back up the data with an understanding of the file system, and inform the FS top revert to normal operation. It may be required that the FS still needs the notifications to be provided by a deamon runing on the host rather than via direct hard disk access, but if the FS is aware and can "advertise" it is in multi access mode, then backup by systems that know the FS in use is quite simple and prevents the host being backed up getting involved in the process much. Chris. ====== Re: New features ====== Chris, This presents several issues that I can see. 1. This scheme is vulnerable to a simple DoS attack. From security point of view, this can be a major vulnerability. Suppose a malicious user manages to write to the particular area indicating that backup is in progress and does not reset the same, there is a major issue. Even if the backup server fails or network is cutoff, then the services provided by the fs are effectively blocked. 2. Maybe, with proper authorization mechanisms, authentication mechanisms you could still authenticate the remote machine However, I am not sure that it is a great idea to stop the file system till the backup server intimates. 3. One thing that can be done is that on receiving the request, the file system calculates the amount of data to be backed up, takes snapshots of the data and pushes it to the server. This would ofcourse be suited to incremental backups. Yogesh. ==== posix and windows acl support ==== I hope to understand the problem after a longer discussion with a samba developer at the Berlin LinuxTag he told me that samba would use windows acl but it can't store it in filesystems which support only posix acl's. windows acl's use more attributes than posix acl's and so the filesystem is the point where the "not posix" attributes must be saved to. Is it possible to implement Posix and Windows acl's in a way that samba can store and edit full User acl's for files and folders? jodbe. ==== Binary prefix, mailing lists and checkout repo ==== There is no way to know if having Binary prefixes http://en.wikipedia.org/wiki/Binary_prefix is going to be there or not. Similarly the mailing list at http://marc.info/?l=linux-fsdevel&r=1&b=200706&w=2 I'm sure is just for the developers, is there something similar like linux-fsdevel-discuss so people like me who wanna know or wanna contribute (but can't code) can do the same. Lastly, where is the repository located? [[User:Shirish|Shirish]] 18:48, 7 June 2007 (UTC)

Return to Talk:New ext4 features.