Previously each request was executed on its own callcontext and
switched to every time the request was being processed. Now requests
borrow the mainloop context and switch only if/when they yield.
This takes two context switches away from a file system request
bringing down the typical amounts 2->0 (e.g. dtfs) and 4->2 (e.g.
psshfs).
The interfaces for manually executing requests changed a bit:
puffs_dispatch_create() and puffs_dispatch_exec() must now be used.
They are not tested, as nothing in-tree wants them and I doubt
anyone else is really interested in them either.
Also do some misc code cleanup related to execution contexts. The
"work-in-progress checkpoint" committed over a year ago was starting
to look slightly weed-infested.
them every time. Speeds up pure in-memory file systems such as
sysctlfs or dtfs quite a bit. For actual I/O-workhorses the result
is of course less tasty.
Get rid of the original puffs_req(3) framework and use puffs_framebuf(3)
instead for file system requests. It has the advantage of being
suitable for transporting a distributed message passing protocol
and therefore us being able to run the file system server on any
host.
Ok, puffs is not quite here yet: libpuffs needs to grow request
routing support and the message contents need to be munged into a
host independent format. Saying which format would be telling,
but it might begin with an X, end in an L and have the 13th character
in the middle. Keep an eye out for the sequels: Parts 3+m/n.
also synchronizes with puffs_mount() and does not return (exit) in the
parent process until the file system has been mounted. This makes
it possible to reliably run e.g. mount_foo jippi /kai ; cd /kai/ee
alternative to the (vastly superior ;) continuation model. This
is very preliminary stuff and not compiled by default (which it
even won't do without some other patches I cannot commit yet).
The raison d'commit of the patch is a snippet which ensures proper
in-order dispatching of all operations, including those which don't
require a response. Previously many of them would be dispatched
simultaneosly, e.g. fsync and reclaim on the same node, which
obviously isn't all that nice for correct operation.
interacts with the userspace file server:
* since the kernel-user communication is not purely request-response
anymore (hasn't been since 2006), try to rename some "request" to
"message". more similar mangling will take place in the future.
* completely rework how messages are allocated. previously most of
them were borrowed from the stack (originally *all* of them),
but now always allocate dynamically. this makes the structure
of the code much cleaner. also makes it possible to fix a
locking order violation. it enables plenty of future enhancements.
* start generalizing the transport interface to be independent of puffs
* move transport interface to read/write instead of ioctl. the
old one had legacy design problems, and besides, ioctl's suck.
implement a very generic version for now; this will be
worked on later hopefully some day reaching "highly optimized".
* implement libpuffs support behind existing library request
interfaces. this will change eventually (I hate those interfaces)
kernel to the file server for silly things the file server did,
e.g. attempting to create a file with size VSIZENOTSET. The file
server can handle these as it chooses, but the default action is
for it to throw its hands in the air and sing "goodbye, cruel world,
it's over, walk on by".
* in addition add/remove, allow enable/disable, which can be used
to control events for descriptors without having to remove all the
data associated with them
* add directsend/receive, which can be used to pass the same buffer
from the caller to read/writeframe and back again
* add flags to enqueue functions and allow urgent buffers to be
processed as the next PDU
equal, larger, respectively instead of 0/1 for non/equal. This
will allow sorting the buffers for faster matching in libpuffs.
While here, change the name from respcmp to framecmp, as that better
reflects the purpose.
NOTE! there is no obvious way to make compilation fail for file
systems which may already be using this feature (although I don't
think there are any outside our tree, as the feature is two weeks
old). Nevertheless, non-updated file systems will fail very quickly.
support removal and addition of i/o file descriptors on the fly.
* detect closed file descriptors
* automatically free waiters of a dead file descriptor
* give the file server the possibility to specify a callback which
notifies of a dead file descriptor
* move loop function to be a property of the mainloop instead of
framebuf (doesn't change effective behaviour)
* add the possibility to configure a timespec parameter which
attempts to call the loop function periodically
* move the event loop functions from the puffs_framebuf namespace
to puffs_framev to differential between pure memory management
functions
stack instead of the continuation stack. This is for lib/36011, where
pthread gets confused since we aren't running on the regular stack.
I'm not really sure which direction to go to with this quite yet, so
make the hack hard to enable on purpose. The whole request dispatch
code needs cleaning anyway.
and event handling mechanisms required in file servers with blocking
I/O backends. puffs_framebuf is built on the concept of puffs_cc
and uses those to multiplex execution where needed.
File systems are required to implement three methods:
* read frame
* write frame
* compare if frame is a response to the given one
Memory management is provided by puffs_framebuf, but the file
systems must still, of course, interpret the protocol and do e.g.
byte order conversion.
As always, puffs_framebuf is work in progress. Current users are
mount_psshfs and mount_9p.
only take the bare essentials, which currently means removing
"maxreqlen" from the argument list (all current callers I'm aware
of set it as 0 anyway). Introduce puffs_init(), which provides a
context for setting various parameters and puffs_domount(), which
can be used to mount the file system. Keep puffs_mount() as a
shortcut for the above two for simple file systems.
Bump development ABI version to 13. After all, it's Friday the 13th.
Watch out! Bad things can happen on Friday the 13th. --No carrier--
accessors for interesting data in it. Namely, you can now get
pu->pu_privdata with puffs_getspecific(), pu->pu_pn_root with
puffs_set/getroot() and pu->pu_maxreqlen with puffs_getmaxreqlen().
system backends which operate purely based on paths, push out more
path management into the library and make path management more
abstract: enable a file system to define a bunch of path management
callbacks, which are used by the framework. Management of normal
/this/is/a/path type paths is provided by the library.
server that it needs to mount the file system backend if it wants
to call mount
* provide some options for getmntopts(), assume that callers will parse
command line (or fstab) args
* reorganize the puffs_cc interface just a bit, preparing for a bigger
revamp later
Add support for having multiple outstanding operations. This is done
by exposing enough interfaces so that it is convenient to have the
main event loop in the implementation itself and by providing a
continuation framework for convinient blocking and rescheduling.
works fine, but will undergo further cleanup & development
pooka