On Mon, 2019-09-23 at 12:08 -0700, Ira Weiny wrote: > Since the last RFC patch set[1] much of the discussion of supporting RDMA with > FS DAX has been around the semantics of the lease mechanism.[2] Within that > thread it was suggested I try and write some documentation and/or tests for the > new mechanism being proposed. I have created a foundation to test lease > functionality within xfstests.[3] This should be close to being accepted. > Before writing additional lease tests, or changing lots of kernel code, this > email presents documentation for the new proposed "layout lease" semantic. > > At Linux Plumbers[4] just over a week ago, I presented the current state of the > patch set and the outstanding issues. Based on the discussion there, well as > follow up emails, I propose the following addition to the fcntl() man page. > > Thank you, > Ira > > [1] https://lkml.org/lkml/2019/8/9/1043 > [2] https://lkml.org/lkml/2019/8/9/1062 > [3] https://www.spinics.net/lists/fstests/msg12620.html > [4] https://linuxplumbersconf.org/event/4/contributions/368/ > > Thank you so much for doing this, Ira. This allows us to debate the user-visible behavior semantics without getting bogged down in the implementation details. More comments below: > <fcntl man page addition> > Layout Leases > ------------- > > Layout (F_LAYOUT) leases are special leases which can be used to control and/or > be informed about the manipulation of the underlying layout of a file. > > A layout is defined as the logical file block -> physical file block mapping > including the file size and sharing of physical blocks among files. Note that > the unwritten state of a block is not considered part of file layout. > > **Read layout lease F_RDLCK | F_LAYOUT** > > Read layout leases can be used to be informed of layout changes by the > system or other users. This lease is similar to the standard read (F_RDLCK) > lease in that any attempt to change the _layout_ of the file will be reported to > the process through the lease break process. But this lease is different > because the file can be opened for write and data can be read and/or written to > the file as long as the underlying layout of the file does not change. > Therefore, the lease is not broken if the file is simply open for write, but > _may_ be broken if an operation such as, truncate(), fallocate() or write() > results in changing the underlying layout. > > **Write layout lease (F_WRLCK | F_LAYOUT)** > > Write Layout leases can be used to break read layout leases to indicate that > the process intends to change the underlying layout lease of the file. > > A process which has taken a write layout lease has exclusive ownership of the > file layout and can modify that layout as long as the lease is held. > Operations which change the layout are allowed by that process. But operations > from other file descriptors which attempt to change the layout will break the > lease through the standard lease break process. The F_LAYOUT flag is used to > indicate a difference between a regular F_WRLCK and F_WRLCK with F_LAYOUT. In > the F_LAYOUT case opens for write do not break the lease. But some operations, > if they change the underlying layout, may. > > The distinction between read layout leases and write layout leases is that > write layout leases can change the layout without breaking the lease within the > owning process. This is useful to guarantee a layout prior to specifying the > unbreakable flag described below. > > The above sounds totally reasonable. You're essentially exposing the behavior of nfsd's layout leases to userland. To be clear, will F_LAYOUT leases work the same way as "normal" leases, wrt signals and timeouts? I do wonder if we're better off not trying to "or" in flags for this, and instead have a separate set of commands (maybe F_RDLAYOUT, F_WRLAYOUT, F_UNLAYOUT). Maybe I'm just bikeshedding though -- I don't feel terribly strongly about it. Also, at least in NFSv4, layouts are handed out for a particular byte range in a file. Should we consider doing this with an API that allows for that in the future? Is this something that would be desirable for your RDMA+DAX use-cases? We could add a new F_SETLEASE variant that takes a struct with a byte range (something like struct flock). > **Unbreakable Layout Leases (F_UNBREAK)** > > In order to support pinning of file pages by direct user space users an > unbreakable flag (F_UNBREAK) can be used to modify the read and write layout > lease. When specified, F_UNBREAK indicates that any user attempting to break > the lease will fail with ETXTBUSY rather than follow the normal breaking > procedure. > > Both read and write layout leases can have the unbreakable flag (F_UNBREAK) > specified. The difference between an unbreakable read layout lease and an > unbreakable write layout lease are that an unbreakable read layout lease is > _not_ exclusive. This means that once a layout is established on a file, > multiple unbreakable read layout leases can be taken by multiple processes and > used to pin the underlying pages of that file. > > Care must therefore be taken to ensure that the layout of the file is as the > user wants prior to using the unbreakable read layout lease. A safe mechanism > to do this would be to take a write layout lease and use fallocate() to set the > layout of the file. The layout lease can then be "downgraded" to unbreakable > read layout as long as no other user broke the write layout lease. > Will userland require any special privileges in order to set an F_UNBREAK lease? This seems like something that could be used for DoS. I assume that these will never time out. How will we deal with the case where something is is squatting on an F_UNBREAK lease and isn't letting it go? Leases are technically "owned" by the file description -- we can't necessarily trace it back to a single task in a threaded program. The kernel task that set the lease may have exited by the time we go looking. Will we be content trying to determine this using /proc/locks+lsof, etc, or will we need something better? > </fcntl man page addition> -- Jeff Layton <jlayton@xxxxxxxxxx>
