kqueue provides a stateful and efficient event notification framework
currently supported events include socket, file, directory, fifo,
pipe, tty and device changes, and monitoring of processes and signals
kqueue is supported by all writable filesystems in NetBSD tree
(with exception of Coda) and all device drivers supporting poll(2)
based on work done by Jonathan Lemon for FreeBSD
initial NetBSD port done by Luke Mewburn and Jason Thorpe
This merge changes the device switch tables from static array to
dynamically generated by config(8).
- All device switches is defined as a constant structure in device drivers.
- The new grammer ``device-major'' is introduced to ``files''.
device-major <prefix> char <num> [block <num>] [<rules>]
- All device major numbers must be listed up in port dependent majors.<arch>
by using this grammer.
- Added the new naming convention.
The name of the device switch must be <prefix>_[bc]devsw for auto-generation
of device switch tables.
- The backward compatibility of loading block/character device
switch by LKM framework is broken. This is necessary to convert
from block/character device major to device name in runtime and vice versa.
- The restriction to assign device major by LKM is completely removed.
We don't need to reserve LKM entries for dynamic loading of device switch.
- In compile time, device major numbers list is packed into the kernel and
the LKM framework will refer it to assign device major number dynamically.
usync_cntl() system calls.
- when usync_cntl is used and the process is aborted (eg: by kill -9)
libc does not call usync_cntl() to unblock things. We have to cleanup
data allocated in the kernel. This is now done through the emulation
specific exit hook
- IRIX initialize some data in the system part of the PRDA: the pid and
a prid (PRDA ID?). We initialize both to pid.
- Move back struct irix_share_group from irix_exec.h to irix_prctl.h, it
is more revelant here.
- fix a few typos
poll will return true until the semaphore is blocked again, but before the
semaphore is blocked, poll returns false.
We do this by maintaining another queue of "released" processes in
struct irix_usema_rec. Unblocking causes the waiting process record to be
moved to the released queue, and poll check for the process in this released
queue.
return the number of processes waiting on the semaphore. We now maintiain
a count of waiting processes.
- Blocked processes are unblocked "first in, first out". We now have a
queue of waiting processes on a asemaphores, so that we can wakeup the
first blocked process.
Problems:
- We now have a lot of dynamic memory allocation, it may be a bit slow.
- Nothing is SMP safe for now. We need to add locks.
- On close, we forget about a semaphore, which is incorrect. One process
can close its fd attached on a semaphore, but other processes would carry
on using it. Since any process can join a shared arena, this is not an
easy thing to solve.
- A lot of usema/usync functionnalities are still to be discovered.
successfully emulates a few test program that use poll semaphores,
including the attach-to-file-descriptor-and-select feature.
There are a few issues:
1) at least one ioctl need to set retval. We handle this in irix_sys_ioctl()
by replacing the data argument by a pointer to a strucutre in the stackgap
that carries the real data and retval. The underlying ioctl methods can
therefore retreive both data and retval.
2) usemaclone is a cloning device: each time it is open, it creates a new
context, and ioctl operation on each open file descriptor will lead to
different behavior. This functionnality is available in NetBSD through the
devvp branch. This first implementation does not use devvp yet, but this
should be done later. Currently, we create a new vnode, and we provide our
own vnode operations. Some operation are applied to the cloned vnode, others
are applied to the original vnode. The v_data field is used to hold a
reference to the original vnode so that we can work on it.
3) at least the setattr vnode operation needs some customisation: IRIX
libc relies on the fact that fchmod on /dev/usema will return 0 in case
of failure.
rafal