Known bugs:

DATE		REPORTED BY	ENTERED BY	MACHINES AFFECTED

03-04-88	hauth/dlb	dbg		RT, SUN
	BSD ptrace interacts badly with processes stopped with
	task_suspend() - if you ptrace() one of these, NO ONE on
	the machine can again use ptrace until it's rebooted.

03-04-88	dbg		dbg		RT, SUN
	thread_set_state won't let user set or clear the trace bit.

03-04-88	sanzi		dbg		RT
	bug in mc881.c that can hang LISP - rich has a fix, but
	it isn't in yet.

03-04-88	jaffe		dbg		SUN
	sd driver loses interrupts, leaving entire machine hanging
	on disk driver.

02-29-88	hauth/jaffe	hauth		RT
	pmap_resident_count() on kernel_map yields a negative number, which
	is then handed off to the user in get_task_info(), which makes ps 
	think the kernel is using negative memory.

02-11-88	hauth/jaffe	hauth		SUN
	panic in pmap_remove_range() with either empty or null pv lists.
	Belived fixed in XM11.

01-28-88	camelot		dbg		VAX, maybe SUN?
	Doing physio (IO to a raw disk) in a multi-threaded task
	causes the pmap to disappear even with wired-down mappings,
	resulting in an addressing fault.

01-25-88	jle		mrt		all
	Running out of paging space generates an error message, but
	that's all. When the page is "paged back in" the process gets
	garbage but no error message.

01-20-88	jle		hauth		RT (XP systems)
	Unstable with lot's o paging

01-20-88	jle		hauth		RT (XM9)
	panic in set_runq() something about paging in a thread while
	killing its task.
	Believed fixed in XM11

01-19-88	jaffe		hauth		sun (XM9)
	Panic in thread_block()

01-19-88	jaffe		hauth		sun (XM9)
	killing & restarting X crashes in winioctl?
	Believed fixed in XM11

12-17-87	mrt		mrt		sun (XM4-7)
	kernel loops printing out SCSI unwanted interrupt on console


real old ones...
	Insufficient locking around utask structures.

	If a thread is created and immediately terminated (before ever running
	at user level), it goes out to a bogus user address.  On the way
	back in, the signal code is checked before thread termination.
	*** buried by new signal/termination check (XM9) - not really fixed
	    on all machines - dbg ***

	During shutdown, the /etc/mach_init process is still running; it
	attempts to page in something that has been terminated by
	vm_object_shutdown.

Things done improperly:
	Copying data on top of "permanent" data in vm_map_copy_entry
	 doesn't do the right thing -- it replaces the address space
	 region, rather than copying.

	vm_map_copy of a region to itself should mark the data copy-on-write

	task_create can't be used, since threads within it will follow
	null proc pointers in clock-interrupt routines (and others),
	crashing the kernel.

Organization problems:

	Thread termination changes for the Sequent and Multimax are not
	yet in place.

	Kernel include files still contain conditionals that should be
	taken from features.h for standard interface functions.
	A (potentially incomplete) list:
		errno.h (RFS)
		file.h (NOPAUSE, RFS, VICE)
		inode.h (RFS?)
		ioctl.h (GENERIC, RPAUSE, SOCKET, TTY, TTYLOC)
		param.h (BUGFIX?)
		ptrace.h (CPUTYPES)
		queue.h (CPUTYPES -- unused?)
		reboot.h (KDB)
		resource.h (GENERIC, RPAUSE)
		table.h (NOFLOAT)
		time.h (GENERIC)
		wait.h (GENERIC)

	Should fix all of the include file references to: be relative to
	the kernel source base directory (i.e., not using "../");
	to use <foo/bar> rather than "foo/bar"; to use one options file.
	Also, should fix all of the copyright headers.
