2000-09-08  Andrey Savochkin  <saw@saw.sw.com.sg>

	* /home/cvs/kernel-patches/2.4.0test7/user_beancounter-IV-0010:
	 - setublimit() syscall interface change to be more flexible and allow
	   more easy adjustments in the future.  The new prototype is
	   source-incompatible with the previous.
	 - select_worst_task() is added for !defined(CONFIG_USER_RESOURCE) case
	   (otherwise the kernel doesn't compile without this option).
	 - proc output stuff is moved to a separate file fs/proc/proc_ub.c
	 - fs/locks.c:
	   - charged_bc was assigned without getting a reference to the
	     beancounter in locks_alloc_lock();  I didn't see immediate
	     problems resulting from it, but it's not robust and a bad style;
	     fixed;
	   - charged_bc reassignment in locks_copy_lock is a bug, fixed;
	   - an extra argument for locks_free_lock() was eliminated and the
	     uncharging policy was clarified: charged_bc is not NULL iff the
	     beancounter should be uncharged;
	   - accounting in new 64-bit lock calls.
	 - beancounter reference leakage (in sys_accept) was fixed.
	 - PTY: fixes to apply for test[67] kernels and eliminate
	   charging/uncharging misbalance.

2000-08-27  Andrey Savochkin  <saw@saw.sw.com.sg>

	* /home/cvs/kernel-patches/2.4.0test7/user_beancounter-IV-0009:
	Port of IV-0008 patch to 2.4.0test7.
	Missing things are
	 - pty limits;
	 - 64-bit file lock calls.

2000-08-25  Andrey Savochkin  <saw@saw.sw.com.sg>

	* /home/cvs/kernel-patches/2.4.0test6/user_beancounter-IV-0008:
	 - Port to 2.4.0test6.
	 - Fix for a bug: null pointer dereference in setublimit() call if it's
	   done before setluid().
	 - Fix for truncated /proc output bug.
	 - Warning isn't printed if beancounter is put with nonzero RSS held.
	 - A bug was spotted in PTY control code.  The code is temporary
	   reverted back to non-beancounter state.

2000-07-31  Andrey Savochkin  <saw@saw.sw.com.sg>

	* /home/cvs/kernel-patches/2.4.0test1/user_beancounter-IV-0007:
	 - accounting for flock and siginfo structures
	 - accounting for number of used pty
	 - tracking the maximum number of held resources over the whole
	   lifetime
	 - /proc/user_beancounter now gives nonprivileged user access to the
	   information about his own beancounter, and only his
	 - new method of accounting of shared pages via a special kernel thread
	 - #ifdef __KERNEL__ is moved in beancounter.h to give userspace the
	   access to resource definitions
	 - UB_MAXVALUE was declared incorrectly, fixed
	 - beancounter referencing from sockets leaked, fixed

2000-07-17  Andrey Savochkin  <saw@saw.sw.com.sg>

	* /home/cvs/kernel-patches/2.4.0test1/user_beancounter-IV-0006:
	Socket accounting, initial version.
	Accounts rcv and snd buffers for AF_INET sockets.

2000-07-06  Andrey Savochkin  <saw@saw.sw.com.sg>

	* /home/cvs/kernel-patches/2.4.0test1/user_beancounter-IV-0005:
	Initial attempt to allow memory use policies:
	 - resident size guarantees;
	 - limits for IPC and anonymous shared memory,
	   small mainstread kernel bug in mm/mmap.c (MAP_SHARED case) is fixed;
	 - allocation guarantees (mmap/brk success);
	 - guarantees against kills because of out-of-memory.

2000-06-27  Andrey Savochkin  <saw@saw.sw.com.sg>

	* /home/cvs/kernel-patches/2.4.0test1/user_beancounter-IV-0004:
	It contains just some bugfixes, namely:
	 - incorrect memory accounting for CLONE_VM-cloned tasks;
	 - a small mistake in RSS accounting;
	 - mremap() memory accounting leak;
	 - cosmetic fix for always-false `if' statement;
	 - Configure.help correction (thanks to Keith Owens <kaos@ocs.com.au>
	   reporting the inaccuracy).
	See http://www.uwsg.indiana.edu/hypermail/linux/kernel/0006.2/1288.html

2000-06-16  Andrey Savochkin  <saw@saw.sw.com.sg>

	* /home/cvs/kernel-patches/2.4.0test1/user_beancounter-IV-0002:
	First port of the old 2.2/2.3 code to 2.4.0test1 kernel.