(no commit message)
[ikiwiki.git] / diary.mdwn
1## Fri 03 August 2007
3* **NATA is now the default ata driver.**
4* **Major changes made to the disk abstraction layer.**
b6af6399 5* **Significant changes made to disk device minor numbers in /dev.**
6* **GPT and 64 bit disklabel support added (non booting).**
7* **The LWP (Userland threading work) is now essentially done.**
b6af6399 8* **Major work on vinum to remove its roll-your-own disklabel layer and to remove its assumptions about the device layer's DMA capabilities.**
9* **Introduce libpthread as a softlink allowing libthread_xu or libc_r to be selected. All multithreaded programs are now linked against libpthread. libc_r is the default as there are still a few issues that need to be worked out with libthread_xu.**
27103d32 10* **Major improvements to wireless networking drivers.**
11* **Many new USB and PCI devices added.**
12* **Numerous new PCI drivers added, particularly w/ networking.**
13* Userland syslink API is now finished.
14* Link state changes are now reported through the routing socket.
15* Signal and proper stop/continue support added to LWPs
16* Entire kernel is now LWP and thread centric instead of process
17 and thread centric.
18* Bind 9.3.4 integrated
19* GCC 4.1.2 integrated, selectable with CCVER. GCC-3 is still
20 the default.
21* Add an abstraction layer to vn_strategy() to hide DMA limitations
22 of underlying devices.
23* Fix symbol resolution issues with loaded modules.
24* Both GCC-3 and GCC-4 are now built by default.
25* Improve virtual kernel disk and networking support. Multiple
26 disks and network interfaces are now supported.
27* Implement SMP support for virtual kernels (mainly as a testbed).
28* Many documentation updates.
29* Enhance kinfo_proc to return LWP info.
30* Enhance the virtual kernel's console.
31* Replace gnu tar with BSD tar.
32* Fix a single stepping bug that could lead to a panic.
33* Allow multiple kernels to be supplied on the release ISO and
34 supply a VKERNEL kernel in addition to the normal kernel.
35* Get rid of the UAREA for good.
36* Synchronize zoneinfo changes with recent standards changes.
37* Fix a mbuf leak in sbappendcontrol().
38* Fix an exec*() syscall memory leak that can lock a machine up
39 when the exec temporary argument space becomes exhausted.
40* Bring in the latest libarchive, bsdtar, sendmail, awk,
41 ping, file, less, and openssh.
42* Bring in binutils 2.17.
43* Support MBs that do not have keyboard controllers by placing
44 an upper limit on the keyboard controller probe timeout.
45* Remove IPV6 type 0 router header support entirely.
46* Implement SYSREF for the kernel - this structure and API
47 implements MP-safe reference counting and also associates a
48 unique 64 bit id with the structure allowing the structure to
49 be referenced off-machine. SYSREF is part of the clustering work.
50* Use SYSREF for struct vmspace and struct vnode, including doing
51 a major revamping of the vnode reference handling code.
52* Do a major reordering of many SYSINITs and a major reordering
53 of where device configuration occurs in the boot sequence.
54 Device configuration now occurs far later in the boot sequence
55 and has access to most major subsystems including threading
56 and major memory subsystems. Interrupts are now enabled during
57 device configuration. This is somewhat experimental but works
58 great in allowing us to locate ordering issues in various device
59 init functions related to their interrupts.
60* Implement asynchronous device attachment but default to
61 disabled for 1.10.
62* Reformulate disk management layer ioctls and remove data
63 pollution. Do not store filesystem-specific data such as
64 the fragment size in the disklabel any more.
65* Expand the SCSI subsystem to handle 16-byte commands,
66 allowing > 32 bit block addressing for devices that support
67 it.
68* Change the 'make upgrade' target to attempt to rebuild the
69 devices found in /dev instead of just rebuilding base devices.
70* Distributions now include a Makefile in /usr with easy-to-use
71 targets for creating and synchronizing /usr/src and /usr/pkgsrc.
72* Greatly simplify LWKT messaging within the kernel.
73* Add virtual CDROM support to the virtual kernel.
74* Add halt and reboot support to the virtual kernel.
75* Fix a bug in bus_dmamap_load() that was not properly following
76 boundary and maximum segment size limits.
77* Save and restore a USB keyboard's translation mode, allowing
78 the keyboard to be un-plugged and re-plugged.
79* Remove the last source of major SMP TLB invalidations in the
80 fork/exec/exit/wait critical paths, significantly reducing IPI
81 traffic and improving fork/exec performance.
82* ktrdump now has an option to loop waiting for new input.
83* Fix mouse jumpiness when emergency interrupts are enabled by
84 not polling PS/2 mouse interrupts.
85* Fix numerous CAM/USB issues so we do not panic or crash if
86 a USB mass storage device is pulled while still mounted. Make
87 sure umount -f works on pulled devices.
88* Synchronize the sound subsystem with FreeBSD.
89* Timeout positive hits in the namecache as well as negative hits
90 for NFS mounts. Otherwise rename-over's can result in a stale
91 namecache that doesn't auto recover.
92* Add MPLS support to traceroute and add a protocol entry in
93 /etc/protocols for MPLS-in-ip.
94* Bring in a number of math library fixes from FreeBSD.
95* **dntpd now properly handles DNS failures, supports DNS pools,
96 and does a quorum consistency check when told to use more then
97 one source.**
99* Fix a bug in the checkpointing code that was preventing a
100 program from being re-checkpointed properly.
101* Major USB code cleanup.
102* Adjust dynamic objcache sizing to handle certain degenerate
103 conditions.
104* Synchronize with the Adaptec SCSI code from FreeBSD.
105* Retire the original ports packaging tools.
108## Tue 30 January 2007
110* **Implement Virtual Kernel support and add a VKERNEL target.**
111 Virtual kernels are run as userland processes and provide a
112 complete (minus hardware drivers) working kernel environment.
113 A simulated disk driver is included and a simulated network
114 driver using the TAP interface is included, allowing network
115 access. See the vkernel(7) manual page for more information.
116* Introduce getcontext(3), setcontext(3), and friends
117* **Release 1.8.0!**
120## Mon 1 January 2007
122* Jails are now IPv6 aware and support multiple IP addresses.
123* Bridging support has been greatly improved.
124* Many, Many new network drivers and chipsets have been added.
125* Major infrastructure for 802.x wireless support added (WPA,
126 TX rate control, major API layer for all wireless network
127 devices, better ifconfig integration, and more).
128* The cpdup program has been greatly enhanced and now supports
129 remote copies via ssh..
130* Improved support for old-style disklabels.
131* Kernel sources are now included on release CDs.
132* Major rewrite of the vnode operations infrastructure, greatly
133 simplifying the layer.
134* Fix a bug related to POSIX locking.
135* Tons of documentation updates and code readability cleanups.
136* Restructure the kernel build to accomodate multiple architectures.
137 Start separting platform components from cpu components.
138* Synchronize a bunch of AMD64 related items (but we don't support
139 64 bit mode yet).
140* Major rewrite of the mount glue infrastructure and NULLFS which
141 allows the namecache topology to be shared for multiple NULLFS
142 mounts and also allows arbitrary stacking of read-only and
143 read-write NULLFS mounts with basically zero kernel overhead.
144* Unconditionally associate a namecache reference with all
145 file descriptors, allowing us to share vnodes across different
146 mounts (NULLFS).
147* NULLFS mounts may now be arbitrarily stacked and distinct paths
148 are no longer required.
149* Properly dereference mount point roots when unmounting. Raise
150 hell if unmounting a filesystem which still has references.
151* Do a massive reorganization of the device operations vector and
152 remove many instances where devops functions were dependant on
153 the calling process or thread to obtain cred information.
154* Do a run through of all system header files to make sure they
155 include any dependancies themselves, instead of requiring the
156 originating source file to include the dependancies first.
157 This greatly reduces the effort required to use kernel header
158 files in new sources.
159* Synchronize MII support with NetBSD and OpenBSD, including adding
160 standard conforming GMII support. Er. In otherwords, improve
161 driver support for physical network interfaces (phys's).
162* Remove VOP locking vectors. Make VOP locking functions direct
163 calls (and thus mandatory).
164* Control access to vnode ref count fields with a spinlock
165* Fix a dubious construct in usr.bin/tail (no, really! That's
166 what the commit message said!).
167* Add support for %j to the kernel printf.
168* Major symbol separation for most kernel functions which
169 inexactly mimic libc functions. e.g. printf -> kprintf,
170 in order to allow user-mode kernels to be linked against libc.
171* Update to sendmail to 8.13.8.
172* Update to OpenSSL to 0.9.8d
173* Update to OpenSSH 4.5p1
174* Update to Bind 9.3.2-P1
175* Update to less 3.9.4
176* Update to awk 20050424
177* Update to file 4.19
178* Update to tcpdump 3.9.5
179* Update to libpcap 0.9.5
180* Synchronize m4 with FreeBSD
181* Import libarchive
182* Import bsdtar 1.3.1 and make bsdtar our default tar. GNU tar
183 is still installed as gtar.
184* Import NetBSD's ftp client under its new name (tnftp).
185* Bring in GCC-4.1 (setenv CCVER GCC41 to use).
186* Significant code cleanups to make the world and kernel compile
187 under GCC-4.1. Note that the default compiler is still GCC 3.4
188 (GCC-4.x will probably become the default sometime in 2007).
189* More work on LWP/THREAD/PROC separation for 1:1 threading support.
190* Fix a major bug in /bin/make that can cause make (or some child
191 make in a large build) to ignore ^C. make temporarily installs
192 SIG_IGN when polling whether a signal is being ignored or not.
193 Fix the code to not do this.
194* Fix a long standing signal/fork race that could cause a process
195 group signal to not make it to a newly fork child if it occurs
196 while the fork is in progress.
197* Synchronize the zoneinfo database with tzdata2006p.
198* Add an ECC detection device, currently supporting AMD64's memory
199 controller.
200* Greatly reduce the memory allocated by fsck when fscking
201 filesytems with a huge number of directories (primarily mirors
202 with lots of hardlinked files). Otherwise fsck can run out
203 of memory on such filesystems.
204* Adjust RCNG to support 'blah=YES/NO' as well as
205 'blah_enable=YES/NO', giving us better compatibility with
206 PkgSrc based RC scripts.
207* Fix a bug related to ALTQ when setting the algorithm for a queue
208 on which packets are already present.
209* Do a major clean up of the BUSDMA header file architecture.
210* We have a new web site layout!
211* Use spinlocks in the objcache instead of tokens (the critical
212 path is still lockless).
213* Replace the global VM page hash table with a per-VM-object
214 Red-Black tree.
215* A considerable amount of work on IPSEC support has been done.
216* NATA has been ported from FreeBSD and is currently being tested.
217* Sync USB support with FreeBSD6 - use task queues to handle
218 operations that cannot be handled from an interrupt thread.
219* Explore low-speed USB busses during cold boot so the USB keyboard
220 starts working earlier in the boot sequence.
224* Start introducing glue for the SYSLINK and CCMS infrastructures.
225 (system link protocol and cache coherency management subsystems).
226* Introduce a system call skeleton and test code for SYSLINK.
227* Introduce the algorithm and structural topology that will be
228 used for cache coherency into the VFS path.
229* Introduce a user-managed virtuallized page table infrastructure
230 which can be accessed via mmap(), as part of the infrastructure
231 to support virtual kernels running in userland (MAP_VPAGETABLE).
232* Introduce vmspace_*() system calls which allows a user process
233 to manage and control multiple VM spaces, as part of the
234 virtualization support effort.
239## Wed 12 July 2006
241* Continued work on LWP/PROC separation.
242* Continued pkgsrc integration.
243* Fix refcount bugs in the kernel module loader and unloader.
244* Lots of netif and serializer cleanups and fixes.
245* Lots of softupdates, filesystem, and buffer cache related fixes.
246* Remove more of the old ports-related infrastructure.
247* Major documentation cleanups.
248* Major code cleanups, ansification.
249* Change the system #include topology so that each include file
250 is responsible for #include'ing any dependant include files.
251* Fix a bug in the PF fragment cache.
252* Random number generator: Instead of generating entropy from
253 selected interrupts (and none by default), we now generate
254 entropy from all interrupts by default and rate limit it to
255 not interfere with high performance interrupts. Completely
256* Random number generator: Completely replace the algorithms,
257 remove limitations on returned bytes from /dev/random (which
258 only served to cause programs to not use /dev/random due to its
259 lack of dependability). Add the ability to seed the RNG.
260 Do some automatic initial seeding at boot.
261* Adjust ssh to find the pkgsrc X11BASE instead of the old ports
262 X11BASE.
263* Fix some compatibility issues in /bin/sh.
264* Fix a small number of critical section enter/exit mismatches.
265* Bring in a bunch of new malloc() features from OpenBSD
266 (guard pages, free page protection, pointer guard, etc).
267* Clean up the DragonFly build system's automatic .patch handling.
268* Bring in openssh 4.3p2.
269* Retire libmsun. It was replaced by NetBSD's libm.
270* Fix a bug in the NFS timer/retry code.
271* Fix issues related to wide-char support.
272* <B>Fix a number of private TSS bugs related to threaded programs.</B>
274* <B>Completely rewrite the user process scheduler and usched APIs.</B>
275* <B>Add system calls that allow a blocking/non-blocking flag to be
276 passed independant of the O_NONBLOCK state of the descriptor.</B>
277* <B>Remove all fcntl(... O_NONBLOCK) calls from libc_r, use the
278 new system calls instead. This solves numerous problems with
279 file descriptors shared between threaded and non-threaded
280 programs getting their non-blocking flag set, and then blowing
281 up the non-threaded program.</B>
282* Add additional red-black (RB) tree function support for ranged
283 searches.
284* Get rid of gdb -k, replace with a separately built kgdb, and
285 build a separate libgdb as well.
286* Implement a VM load heuristic. Remove the full-process SWAP
287 code which never worked well and replace with page-fault
288 rate-limiting code based on the VM load.
289* Fix a serious bug in adjtime()'s microseconds calculation.
290* Fix a serious bug in libc's strnstr() function. The function
291 was testing one byte beyond the specified length limit. This
292 can cause a seg-fault when, e.g. using strnstr() on a memory
293 mapped file whos last byte abuts the end of the VM page.
294* Bring in sendmail 8.13.7.
295* Bring in SHA256 support from FreeBSD.
296* Implement a hardlink-mirroring option (-H) in cpdup.
297* Add missing code needed to detect IPSEC packet replays.
298* Enable TCP wrappers in sshd.
299* Restrict recursive DNS queries to localhost by default.
300* <B>Massive reorganization and rewrite of the 802_11 subsystem,
301 with many pieces taken from FreeBSD.</B>
303* Fix a number of issues with user-supplied directory offsets
304 blowing up readdir/getdirentries for NFS and UFS.
305* Normalize internal kernel I/O byte lengths to 'int' and remove
306 a ton of crazy casts to and from unsigned int or size_t.
307* Remove NQNFS support. The mechanisms are too crude to co-exist
308 with upcoming cache coherency management work and the original
309 implementation hacked up the NFS code pretty severely.
311 They never lived up to their billing and DragonFly doesn't
312 need them since DragonFly is capable of aliasing vnodes via
313 the namecache. Rearrange the VFS code such that VOP_OPEN is
314 now responsible for assocating a VM object with a vnode.
315* Formalize open/close counting for the vnode and assert that
316 they are correct.
317* Remove the thread_t argument from most VOP and VFS calls.
318 Remove the thread_t argument from many other kernel procedures.
319* Integrate FreeBSD's new ifconfig(8) utility.
320* Fix a race condition in the floating point fault code that
321 could sometimes result in a kernel assertion.
322* Fix a crash in the TCP code when the MTU is too small to
323 support required TCP/IP/IPSEC options and headers.
324* Separate EXT2 conditionals from the UFS code, copying the files
325 in question to the EXT2 directory instead of trying to
326 conditionalize them. Also remove function hooks and other code
327 mess that had been implemented to allow the UFS code to be
328 used by EXT2.
329* Greatly simplify the lockmgr() API. Remove LK_DRAIN,
330 LK_INTERLOCK, and many other flags. Remove the interlock
331 argument.
332* Fix a bug in the POSIX locking code (lockf). Actually, completely
333 rewrite the POSIX locking code. The previous code was too
334 complex and mostly unreadable.
335* Do a major clean up of all *read*() and *write*() system calls,
336 and iovec handling.
337* Replace many instances where LWKT tokens are used with spinlocks.
338* Make spinlocks panic-friendly. Properly handle the detection
339 of indefinite waits and other deadlock issues.
340* Improve network performance by embedding the netmsg directly in
341 the mbuf instead of allocating a separate netmsg structure for
342 each packet.
343* <B>Implement both shared and exclusive spinlocks.
344 Implement a much faster shared spinlock. Cache the shared
345 state such that no locked bus cycle operation is required in
346 the common case.</B>
348* Implement msleep(). Use a novel approach to handling the
349 interlock that greatly improves performance over other
350 implementations.
351* Add cpu-binding support to the scheduler and add a system call
352 to access it. A user process can be bound to a subset of cpus.
353* <B>Prefix all syscall functions in the kernel with 'sys_'
354 to reduce function prototype pollution and incompatibilities,
355 and to eventually support virtualized kernels running in
356 userland.</B>
357* Port the enhanced SpeedStep driver (EST) for cpu frequency control.
358* Remove the asynchronous syscall interface. It was an idea before
359 its time. However, keep the formalization of the syscall
360 arguments structures.
361* Add a facility which statistically records and dumps program
362 counter tracking data for the kernel.
363* Improve kernel SSE2-based bcopies by calling fnclex instead of
364 fninit.
365* Major BUF/BIO work - <B>make the entire BUF/BIO subsystem BIO
366 centric.</B>
367* Major BUF/BIO work - <B>get rid of block numbers, use 64 bit
368 byte offsets only.</B>
370* Major BUF/BIO work - Clean up structures and compartmentalize
371 driver-specific private fields. Rewrite and simplify device
372 and vnode strategy APIs.
373* Major BUF/BIO work - Remove B_PHYS. There is no longer any
374 differentiation between physical and non-physical I/O at
375 the strategy layer.
376* Major BUF/BIO work - Replace the global buffer cache hash table
377 with a per-vnode RB tree. Add sanity checks. Require that all
378 vnode-based buffers be VMIO backed.
379* MPSAFE work - <B>Implement the parallel route table algorith.</B>
380* MPSAFE work - <B>Make the user process scheduler MPSAFE.</B>
381* MPSAFE work - <B>File descriptor access is now MPSAFE. Many
382 fd related functions, like dup(), close(), etc, are either MPSAFE
383 or mostly MPSAFE.</B>
384* MPSAFE work - <B>Push the BGL deeper into the kernel call stack
385 for many system calls.</B>
387* MPSAFE work - <B>Make the process list MPSAFE.</B>
388* MPSAFE work - <B>Make all cred functions MPSAFE.</B>
389* NRELEASE - compilable kernel sources are now included on the ISO.
392## Sat 7 January 2006
394* Add the closefrom() system call.
395* GCC 3.4 is now the default compiler. 2.95.x is no longer
396 supported (it can't handle the new threading storage classes
397 properly).
398* Import Citrus from NetBSD.
399* Implement direct TLS support for programs, whether threaded or not.
400* Major library and user-visible system structure changes
401 (dirent, stat, errno, etc), and other work requires a major
402 library bump for libc and other libraries. libc is now
403 libc.so.6.
404* stat: inode size now 64 bits, nlink now 32 bits. new fields,
405 added pad.
406* dirent: inode size now 64 bits, various fields disentangled from
407 the UFS dirent.
408* statfs: new fields, added pad.
409* Clean up RC scripts that are not used by DragonFly.
410* Remove the OS keyword requirement for RC scripts.
411* Add support for unsigned quads to sysctl.
412* Implement DNTPD, DragonFly's own NTP client time synchronization
413 demon.
414* Correct a large number of bugs in the third party ntpd code, but
415 for client-side operations we now recommend you use dntpd.
416* Add a framework for aggregating per-cpu structures for user
417 reporting.
418* Userland TLS (data spaces for threads) support added.
419* Create a binary library compatibility infrastructure that
420 allows us to install and/or upgrade older revs of shared
421 libraries on newer machines to maintain compatibility with
422 older programs.
423* Fix issues related to the expansion of symbolic links by the
424 bourne shell.
425* Many, many mdoc cleanups and fixes.
426* Update cvs, openssl, ssh, sendmail, groff,
427 and other numerous contributed applications.
428* Bring in a brand new PAM infrastructure.
429* Introduce pkgsrc support.
430* Get rid of libmsun.
431* Implement backwards scanning and partial-transaction handling
432 features in jscan.
434* FreeBSD-SA-05:06.iir - major disk access vulnerability for IIR
435* FreeBSD-SA-05:04.ifconf - memory disclosure vulnerability
436* FreeBSD-SA-05:08.kmem - memory disclosure vulnerability
437* FreeBSD-SA-05:16.zlib - possible buffer overflow in zlib
438* FreeBSD-SA-05:18.zlib - possible buffer overflow in zlib
439* FreeBSD-SA-05:15.tcp - fix TCP RESET window check
440 (DOS attack vulnerability)
441* ? - a bzip2 vulnerability
443* Fix a bug in the TCP NewReno algorithm which could result in
444 a large amount of data being unnecessarily retransmitted.
445* Fix numerous TCP buffering issues.
446* Implement TCP Appropriate Byte Counting
447* Bring in ALTQ and reorganize the IF queueing code to remove
448 per-driver dependencies on ALTQ.
449* Strip away numerous TCP hidden indirections that make code hard
450 to read and understand.
451* Introduce BPF_MTAP which includes an address family parameter.
452* Reimplement network polling with a systimer, allowing the
453 frequency to be adjusted on the fly.
454* Remove the really bad hack that was calling the network polling
455 code from the trap code.
456* Completely rewrite network polling support.
457* Make the network IF serializer mandatory for all network device
458 driver interrupts, ioctl's, and if_ callbacks.
459* Implement a very fast memory object caching infrastructure. This
460 will eventually replace zalloc() (but not yet).
461* Rewrite the mbuf allocator using the new memory object caching
462 infrastructure. Remove many crazily-large mbuf macros in favor
463 of the new infrastructure.
464* Convert all remaining uses of the old mbuf m_ext API to the new
465 API. Remove support for the old API.
466* Reorder the detach sequence in all network drivers. Unhook the
467 interrupt first rather then last.
468* Fix all instances where an mbuf packet header and mbuf data
469 buffer were being referenced by the wrong name and all instances
470 where the packet header flag was being improperly set or cleared.
471* Fix a number of mbuf statistics counting bugs.
472* Fix numerous bugs in ipfw/ipfw2 where m_tag data was not being
473 stored in the right place, resulting in a panic.
474* Add support for the experimental SCTP protocol.
475* Fix an issue with cloned interfaces being added twice.
477* Add a passive IPIQ call for non-time-critical events such as
478 memory free() calls.
479* Add TLS support for threads using the GDT instead of the LDT.
480* Greatly simplify and demystify the NTP kernel interface. Convert
481 most aspects of the interface over to sysctls.
482* Implement ranged fsync's in-kernel. This capability will
483 eventually replace the write-behind heuristic.
485* Introduce MP-safe mountlist scanning code.
486* Introduce rip-out-safe red-black tree scanning code.
487* Use the new RB scanning code to get rid of VPLACEMARKER and
488 generally use the new RB scanning code to handle all RB tree
489 scanning in a safe way (allowing the scan code callback to block).
490* Zoneinfo upgrades
491* Rename cpu_mb*() functions to cpu_mfence(), cpu_lfence(), and
492 cpu_sfence() to make their function more apparent.
493* Fix bugs in the LWKT token code related to token references
494 being lost due to a preemption or blocking condition.
495* Fix bugs in the LWKT rwlock code relating to preemption occuring
496 during token acquisition.
497* Fix a bug in the LWKT thread queueing and dequeueing code
498 related to a preemption.
499* Increase the size of the physmap[] array to accomodate newer
500 PCs which have a larger number of memory segments and fix
501 an overflow bug.
502* Use the ACPI timer if present instead of one of the other 8254
503 timers (which are not dependable because BIOS calls might
504 manipulate them).
505* Change cpu statistics to be accounted for on a per-cpu basis.
506* Make network routing statistics per-cpu.
507* Extend the interrupt vector code to pass a frame as a pointer.
508* Remove the last vestiges of the old mbuf tagging code.
509* Add a serializer API and code (basically blockable mutexes).
510* Add interrupt enablement and disablement features to the new
511 serializer module to deal with races against blocked serializer
512 locks when e.g. removing a driver.
513* Remove bus_{disable,enable}_intr(), it was not generic enough
514 for our needs.
515* Remove all spl*() procedures and convert all uses to critical
516 sections.
517* Do not try to completely halt all cpus when panic()ing as this
518 will likely leave the machine in a state that prevents it from
519 being able to do a dump.
520* Try to unwind certain conditions when panic()ing from a trap
521 in order to give the machine a better chance to dump its core.
522* A number of malloc()'s using M_NOWAIT really needed to be
523 using M_WAITOK.
524* Attempt to avoid a livelocked USB interrupt during boot by
525 delaying the enablement of the EHCI interrupt until after all
526 companion controllers have been attached.
527* Reimplement the kernel tracepoint facility (KTR) to greatly
528 reduce the complexity of the API as well as remove all hardwired
529 flags and values. In addition, record two levels of call
530 backtrace for each entry, if enabled.
531* Beef up ktrdump to display symbolic results when possible.
532* Beef up the slab allocator build with INVARINTS by adding a
533 bitmap to detect duplicate frees and such.
534* Remove the 16 bit count limit for file descriptors.
535* Replace the file descriptor allocator with an O(log N)
536 full-on in-place binary search tree.
537* Allow the initial stack pointer for a use process to be
538 randomized.
539* Fix numerous scheduling issues that could cause the scheduler
540 to lose track of a reschedule request, resulting in poor
541 interactive performance. Rewrite the interactive/batch
542 heuristic.
543* Begin to implement a management system to allow multiple
544 userland schedulers to be configured in a system.
545* Add rm -I and add an alias for interactive shells to use it
546 by default. -I is a less invasive -i.
547* Fix a bug in the pipe code that was not handling kernel-space
548 writes correctly. Such writes can occur whenever the kernel
549 writes KVM-referenced data to a descriptor, such as that
550 journaling code might do.
551* Fix many issues with the high level filesystem journaling code.
552 High level journal records are now considered fairly solid.
553* Implement the transactional features of the high level journaling
554 subsystem by allowing a journaling record to be written prior to
555 the VFS operation being executed, then aborted if the VFS operation
556 fails.
557* Implement UNDO records for most journaling transaction types.
558* Implement the journaling code's full-duplex ack protocol feature
559 which allows journals to be broken and restarted without losing
560 data.
561* Implement a stat-visible FSMID (filesystem modification id). This
562 identifier changes whenever any modifying operation on the file
563 or directory occurs, and for directories this identifier also
564 changes if anything in the sub-tree under the directory is
565 modified (recursively). The FSMID is synthesized for filesystems
566 which do not implement it directly in order to guarantee its
567 usefulness for at least a subset of operations.
568* Implement pesistent storage of the FSMID for UFS.
569* Implement shutdown() support for pipes.
570* Implement a low level spinlock facility. Basically the
571 implementation gives us an MP-safe critical section type of
572 vehicle. However, being a spinlock the facility may only be
573 used for very short sections of code.
574* Fix a bug with USB<->CAM communication for USB mass storage
575 devices.
576* Fix numerous bugs in USB, primarily EHCI.
577* Fix multiple panics when a fatal trap occurs from an IPI or
578 FAST interrupt. Interlock panics on multiple cpus so only the
579 first is recognized as the 'real' panic.
580* Add a large number of assertions to the scheduler and interrupt
581 subsystems.
582* Fix a critical IPI messaging bug (SMP only).
583* Do not compile the kernel with the stack protector. The stack
584 protector generates weird incorrect or unexpected code in some
585 cases which interfere with the C<->assembly interactions in the
586 kernel build
587* Various bug fixes to softupdates.
588* Fix a bitmap scanning bug in UFS which could sometimes result
589 in a sanity check panic, but no data corruption.
590* Fix a deadlock in UFS's ffs_balloc() related to an incorrect
591 buffer locking order.
592* Continued work on the buffer cache.
593* Separate out APIC and ICU interrupt management.
594* Rewrite the interrupt setup code.
595* Major rewriting of the VFS directory scanning code. Add a new
596 function vop_write_dirent() to create the dirent for return to
597 userland. The new API is mandatory and filesystem code (not
598 even UFS) may not make assumptions about the size of the
599 userland-returned dirent.
600* Major cleanup of the device identification method.
601* Lots of driver updates.
602* ANSIfy a great deal more of the codebase.
603* Remove the now obsolete smp_rendezvous() mechanism.
604* Compile up both the TFTP and the NFS PXE bootp code rather
605 then the (previous) make.conf option to select one or the other.
606* Convert the lockmgr interlock from a token to a spinlock, also
607 incidently fixing an issue where non-blocking locks would
608 still potentially issue a thread switch.
609* Fix bugs in the interrupt livelock code.
610* Rewrite the code handling stopped user processes.
611* Rewrite tsleep()/wakeup() to be per-cpu and MPSAFE. Reorganize
612 the process states (p_stat), removing a number of states but
613 resynthesizing them in eproc for 'ps'.
614* Integrate the new if_bridge code from Open/Net/FreeBSD.
615* Add an emergency interrupt polling feature that can be used
616 to get an otherwise non-working system working.
620## Thu 7 April 2005
622* Fix numerous issues in the new namecache code. The new code is
623 now considered to be solid (Matt).
624* Fix numerous issues in the NFS code related to cache timeouts
625 not working properly, truncate operations not working properly
626 in certain circumstances, and TCP disconnections due to
627 races in accessing the streamed data. NFS was also not properly
628 setting TCP_NODELAY which resulted in poor performance over
629 TCP connections (Matt).
630* Split out kernel information access abstraction into libkcore
631 and libkinfo (joerg).
632* Optimize the interrupt processing code a bit (Matt).
633* Significant progress in the journaling layer has been made,
634 but it isn't ready for production use yet. The journaling
635 layer is now capable of encoding a forward journaling stream
636 (the undo data is not yet being encoded). Still TODO: undo
637 data, two-way transaction acknowledgement protocol, and
638 emergency swap backing to allow system operation to proceed
639 in the event of a long-term journaling stream stall. (Matt)
640* Fix numerous issues related to SMP-distributed wildcard
641 listen sockets for TCP. (Jeff, Matt)
642* Major cleanup of the route table code in preparation for
643 SMP replication work (Jeff)
644* ALTQ Integration (Joerg)
645* Fix MTU discovery which appears to have been broken
646 forever (Jeff).
647* Miscellanious driver updates. (Various)
648* Miscellanious ACPI updates.
649* Fix a security issue with sendfile() (Various)
650* Lots of work on the make program (Max)
651* Add TLS support calls to the kernel to create segments suitable
652 for loading into %gs. (Matt, David Xu)
653* Add TLS support to the dynamic linker (David Xu, Joerg)
654* Add kernel support for mmap-based userland mutexes
655* Continuing work on a new threading library (David Xu)
656* Add jail support for varsyms (Joerg)
657* Incorporate the Minix MINED editor into the /bin build for use
658 as an emergency editor while in single-user.
659* Fix an issue with PCMCIA card removal.
660* Replace the VM map lookup data structures with a red-black tree.
661* Fix a potentially serious issue with inodes being reused before
662 their associated vnodes have been entirely recycled. This also
663 prevents unnecessary stalls on recycle flushing against new
664 file creation that would otherwise have to block waiting for
665 the inode it chose to finish cycling.
666* Fix some minor issues with /dev/tty operations.
667* Fix an issue with the vnode recycler. Certain types of file
668 extractions could cause it to stop oprating due to not being
669 able to clean out the namecache topology. This most often
670 occured when extracting archives with a large number of
671 hard links.
672* Fix two issues that caused the system's time of day tracking to
673 either stop operating or operate incorrectly.
674* Enhance the checkpointing code to save and restore the
675 signal mask.
676* Major work on the XIO and MSFBUF infrastructures as we begin to
677 use them in the kernel more. (Hiten, Matt)
678* Implement CLOCK_MONOTONIC for the clock_*() system calls.
679* Fix a bug in the TCP limited transmit code.
680* Fix a crash that sometimes occurs when using the firewire
681 console.
686## Mon 20 December 2004
688* The old timeout() API has been completely ripped out and replaced
689 with the newer callout_*() API.
690* USB support has been synchronized with FreeBSD and NetBSD
691* USB keyboard detachment and reattachment while X is active no
692 longer messes up the translation mode.
693* Fix a keyboard lockup issue (the new callout code was not being
694 properly called in an ATKBD hack to deal with lost interrupts).
695* **TCP SACK support is now considered production stable (Jeff).**
696* A TCP connection closing bug has been fixed, related to changes
697 we've made to the TCPS state values.
698* **Expand TCP's header prediction case to handle common
699 window updates (Jeff), greatly improving cpu efficiency.**
700* **Implement tail-append to sockbufs for the TCP stack, greatly
701 improving cpu efficiencies when dealing with large TCP buffers.**
703* Lots of miscellanious network layer cleanups (Joerg).
704* Fix an IPv6 pcb replication issue that was creating problems
705 with e.g. Apache-2.0.
706* kernel event logging ported from FreeBSD (eirikn).
707* DCONS (console over firewire) support added (simokawa/from FreeBSD)
708* Additional GigE drivers added.
709* Major VFS messaging and interfacing progress. The old namei()
710 and lookup() API has been completely removed. All high level
711 layers now use the new API and run through a compatibility layer
712 to talk to VFSs which still for the most part implement the old
713 VOP calls. Use locked namespaces to protect RENAME, REMOVE,
714 MKDIR, and other calls rather then depending on locked vnodes
715 for those protections.
716* More VFS work. Keep track of the current directory with a
717 namecache pointer rather then a vnode pointer.
718* More VFS work. Rewrite the vnode interlock code during
719 deactivation and disposal. Begin consolidating v_lock.
720* Adjust the boot code to operate more deterministically by always
721 using EDD (linear block number) mode first, only falling back
722 to CHS if EDD fails. Before it would try to use CHS for
723 cylinders < 1024, resulting in non-deterministic operation on
724 modern machines.
725* Separate out loader configuration files for BOOTP vs non-BOOTP
726 boots.
727* IPSEC code moved to netproto/ipsec, update, and cleanups
728 (Pawel Biernacki).
729* Minor fixes to PPP.
730* Performance cleanup of if_em. Fix alignment requirements to
731 reduce instances where bounce buffers are allocated (from FreeBSD).
732* Kernels built with debug info are now installed with debug info,
733 greatly improving normal enduser's ability to provide useable
734 bug reports. The backup copy of the kernel is stripped when
735 the copy is made.
736* **A number of bug fixes to the VM system seem to have fixed the few
737 remaining long-term panics in DragonFly. In particular, a very
738 serious bug in contigmalloc() inherited from FreeBSD-4.x has been
739 fixed We now consider DragonFly as stable as FreeBSD-4.x.**
741* Many minor driver bug fixes here and there, including one to the
742 serial driver which was responsible for machine lockups in
743 certain cases.
744* **Major expansion of the checkpoint code API. You can now
745 re-checkpoint programs that have been checkpoint-restored.
746 The system call for checkpointing and checkpoint restore
747 functions is now official. Certain VM area issues have been
748 fixed. And it is now possible (and easy!) to write
749 checkpoint-aware programs.**
750* Abstract kernel structure access via libkinfo (joerg).
751* Fix a number of timer issues related to the 8254, sleep/wakeup,
752 and recovery from clock jumps due to high latencies.
753* Do better ESTALE checking for NFS clients to reduce instances
754 where ESTALE makes it all the way back to the application layer.
755* Improved polling support for UHCI USB (drhodus).
756* Fix /boot/loader's handling of extended DOS partitions. There was
757 an off-by-one issue that prevented the boot loader from passing
758 the proper slice number to the kernel in certain cases (walt).
759* Integration of PF as third firewall (joerg, corecode, Devon O'Dell,
760 from FreeBSD/OpenBSD)
763## Sat 18 September 2004
765* DragonFly has adopted the 3-clause BSD license as its official
766 license.
767* NFS - increase the size of the nfsheur hash table as per a
768 Freenix track paper. nfsheur is used for sequential I/O
769 heuristic in NFS. The increase greatly improves clustering
770 under parallel NFS loads.
771* More VFS work: make vnode locks mandatory, get rid of nolock
772 kludges in NFS, procfs, nullfs, and many other filesystems.
773 Remove the vnode_if.m dynamic dispatch algorithms and replace
774 with a fixed structure (in preparation for upcoming VFS work).
775 Rewrite the VOP table parsing code. Redo vnode/inode hash table
776 interactions to close race conditions (Matt).
777* Reorganize the boot code to consolidate all fixed ORG directives
778 and other dependancies into a single header file, allowing us
779 to re position the boot code and eventually (not yet) move it
780 out of low BIOS memory. Change the way the BTX code clears BSS
b6af6399 781
782 to make it more deterministic and more gcc3 compatible.
783* A major cleanup of ipfilter has been done (Hiten).
784* Fix a very old but serious bug in the VM system that could result
785 in corrupt user memory when MADV_FREE is used (from Alan Cox).
786* Rewrite most of the MBUF allocator and support code. Get rid
787 of mb_map and back the mbuf allocator with malloc. Get rid of
788 the old-style m_ext support and replace with new style m_ext
789 support (which is somewhat different then FreeBSD's new code).
790 Cleanup sendfile() to use the new m_ext callback scheme.
791* Lots minor/major bug fixes, cleanups, new driver support, etc.
792* More GCC3 work. The system mostly compiles and runs GCC3 builds
793 but the boot/loader code still has some issues (Various people).
794* Work on the userland schedule. Introduce an 'interactivity'
795 measure in an attempt to do a better job assigning time slices.
796 Fix some scheduler interaction bugs which were sometimes resulting
797 in processes being given a full 1/10 second time slice when they
798 shouldn't be.
799* A major import of the FreeBSD-5 802_11 infrastructure has been
800 accomplished (Joerg).
801* NDis has been ported from FreeBSD, giving DragonFly access to
802 many more 802.11 devices via windoz device drivers (Matt).
803* Add thermal control circuit support (Asmodai).
804* Generally add throttling support to the system.
805* **Add TCP SACK support (Jeffrey Hsu). This is still considered
806 experimental.**
808* Make the syncache per-cpu and dispatch syncache timer events
809 via LWKT messages directly to the appropriate protocol thread
810 (Jeffrey Hsu). This removes all race conditions from the syncache
811 code and makes it 95% MP safe.
812* **Greatly reduce the number of ACKs generated on a TCP connection
813 going full-out over a GigE (or other fast) interface by delaying
814 the sending of the ACK until all protocol stack packets have been
815 processed. Since GiGE interfaces tend to aggregate 8-12(+)
816 received packets per interrupt, this can cut the ACK rate by 75%
817 (one ack per 8-12 packets instead of one ack per 2 packets), and
818 it does it without violating the TCP spec. The code takes
819 advantage of the protocol thread abstraction used to process
820 TCP packets. (Matt)**
821* **Greatly reduce the number of pure window updates that occur over
822 a high speed (typ GigE) TCP connection by recognizing that a
823 pure window update is not always necessary when userland has
824 drained the TCP socket buffer. (Jeffrey Hsu)**
825* Rewrite the callout_*() core infrastructure and rip out the old
826 [un]timeout() API (saving ~800K+ of KVM in the process). The new
827 callout infrastructure uses a DragonFly-friendly per-cpu
828 implementation and is able to guarantee that callouts will occur
829 on the same cpu they were registered on, a feature that the TCP
830 protocol stack threads are going to soon take major advantage of.
831* Cleanup the link layer broadcast address, consolidating many
832 separate implementations into one ifnet-based implementation
833 (Joerg).
834* BUF/BIO progress - start working the XIO vm_page mapping code
835 into the system buffer cache (Hiten). Remove b_caller2 and
836 b_driver2 field members from the BUF structure (Hiten) (generally
837 we are trying to remove the non-recursive-friendly driver
838 specific fields from struct buf and friends).
839* The release went well! There were a few gotchas, such as trying
840 to run dual console output to the serial port causing problems on
841 laptops which did not have serial ports. A bug in the installer
842 was serious enough to have to go to an '1.0A' release a day or two
843 after the 1.0 release. But, generally speaking, the release did its
844 job!
845* More USB fixes. Clean up some timer races in USB/CAM interactions
846 related to pulling out USB mass storage cards. Fix a serious bug
847 that could lead to lost transactions and create confusion between
848 the USB code and the device.
849* Async syscall work: clean up the sendsys2() syscall API into
850 something that's a bit more reasonable (Eirik Nygaard)
851* Add a generic framework for IOCTL mapping (Simon).
852* Add VESA mode support, giving us access to bitmapped VESA video
853 modes (Sascha Wildner).
854* Fix USB keyboard support by giving the USB keyboard preference
855 even if a normal keyboard is detected earlier in the boot process.
856 This is necessary due to hardware/firmware level PS/2 keyboard
857 emulation that many USB chipsets and BIOSes offer.
858* **Installer Updated: Lots of bug fixes have been made.**
860* Stability: Spend two weeks stabilizing recent work in preparation
861 for another big push.
864## Sun 11 July 2004
866* Master ISO for the 1.0-RELEASE is now on the FTP site, we
867 release on Monday 12-July-2004.
868* **Major revamping of the boot code. Support dual-console mode
869 (video and serial). Properly detect missing serial ports
870 (common for laptops). Initialize the serial port to 96008N1.**
871* Bring ACPICA5 uptodate and integrate it into the build as a
872 KLD, just like FreeBSD-5. Fix additional issues as well.
873 Nearly all of this work was done by YONETANI Tomokazu. Use
874 the latest INTEL code (20040527). This greatly improves
875 our laptop support.
876* Make the kernel check both CDRom drives for a root filesystem
877 when booted with -C, which allows the boot CD to work in either
878 drive on systems with two drives.
879* **Integrate the DragonFly Installer into the release build. The
880 DragonFly Installer is a from-scratch design by Chris Pressey,
881 Devon O'Dell, Eirik Nygaard, Hiten Pandya & GeekGod (aka
882 Scott Ullrich). The design incorporates a worker backend and
883 multiply targetable frontends. Currently the console frontend
884 is enabled, but there is also a CGI/WWW frontend which is on the
885 CD but still considered highly experimental.**
887* Fix a bug in the polling backoff code for the VR device
888* Implement interrupt livelock detection. When an interrupt
889 livelock is detected the interrupt thread automatically
890 throttles itself to a (sysctl settable) rate. When the interrupt
891 rate drops to half the throttle limit the throttling is removed.
892 This greatly improves debugability, especially on laptops with
893 misrouted interrupts. While it doesn't necessarily fix the
894 broken devices it does tend to allow the system to continue to
895 operate with those devices that *do* still work.
896* Properly probe for the existance of the serial port in the
897 kernel, which allows us to run a getty on ttyd0 by default for
898 the release CD. Prior to this fix attempting to use ttyd0 would
899 result in a system deadlock.
900* Bring in some pccard driver improvements from FreeBSD, including
901 increasing the CIS area buffer from 1K to 4K and properly
902 range-checking the CIS parser (which avoids panics and crashes).
903* Update our DragonFly Copyright and update date specifications
904 on a number of copyrights.
905* Add support for randomized ephermal source ports.
906* Do some network driver cleanups.. basically move some common code
907 out of the driver and into ether_ifattach() (work by Joerg).
908* Try to be more compatible with laptop touchpads whose aux ports
909 return normally illegal values (Eirik Nygaard).
910* Add common functions for computing the ethernet CRC, work
911 by Joerg, taken from NetBSD.
912* Add support for additional AGP bridges - taken from FreeBSD-5.
913* Add support for the 're' network device.
914* Bring OHCI and EHCI up-to-date with NetBSD.
915* Miscellaneous driver fixes to ips, usb (ugen), sound support.
916* Miscellaneous Linux emulation work, by David Rhodus, taken
917 from FreeBSD-SA-04:13.linux.
920## Sun 27 June 2004
922* We are going to release 1.0RC1 tonight.
923* Implement interrupt livelock detection and rate limiting. It's
924 still a bit raw, but it does the job (easily testable by plugging
925 and unplugging cardbus cards at a high rate). The intent is to
926 try to make the system more survivable from interrupt routing and
927 other boot-time configuration issues, and badly written drivers.
928* Implement an emergency polling mode for the VR device if its
929 interrupt does not appear to be working.
930* Implement dual console (screen and serial port) support by
931 default for boot2 and the loader.
934## Mon 21 June 2004
936* Joerg has brought GCC-3.4-20040618 in and it is now hooked up
937 to the build. GCC-3.3 will soon be removed. To use GCC-3.4
938 'setenv CCVER gcc34'.
939* The world and kernel now builds with gcc-3.4. The kernel builds
940 and runs with gcc-3.4 -O3, but this is not an officially supported
941 configuration.
942* More M_NOWAIT -> M_INTWAIT work. Most of the drivers inherited
943 from FreeBSD make aweful assumptions about M_NOWAIT mallocs
944 which cause them to break under DragonFly.
945* /usr/bin/ps now reports system thread startup times as the boot
946 time instead of as Jan-1-1970. The p_start field in pstats has
947 been moved to the thread structure so threads can have a start
948 time in the future - Hiten.
949* Zero the itimers on fork() - Hiten / SUSv3 compliance.
950* MMX/XMM kernel optimizations are now on by default, greatly
951 improving bcopy/bzero/copyin/copyout performance for large (>4K)
952 buffers.
953* A number of revoke() related panics have been fixed, in particular
954 related to 'script' and other pty-using programs.
955* The initial MSFBUF scheme (multi-page cached linear buffers)
956 has been committed and is now used for NFS requests. This is
957 part of the continuing work to eventually make I/O devices
958 responsible for any KVM mappings (because most just set up DMA
959 and don't actually have to make any) - Hiten.
960* Continuing ANSIfication work by several people - Chris Pressey.
961* Continuing work on the LWKT messaging system. A number of bugs
962 in the lwkt_abortmsg() path have been fixed.
963* The load average is now calculated properly. Thread sleeps were
964 not being accounted for properly.
965* Bring in a number of changes from FreeBSD-5: try the elf image
966 activator first.
967* Fix a number of serious ref-counting and ref holding bugs in
968 procfs as part of our use of XIO in procfs - GeekGod and Matt
969* Use network predicates for accept() and connect() (using the
970 new message abort functionality to handle PCATCH) - Jeff.
971* Convert netproto/ns to use the pr_usrreqs structure.
972* Implement markers for traversing PCB lists to fix concurrency
973 problems with sysctl.
974* Implement a lwkt_setcpu() API function which moves a thread
975 to a particular cpu. This will be used by sysctl to iterate
976 across cpus when collecting per-cpu structural information.
977* Redo netstat to properly iterate the pcb's across all cpus.
978* Significant mbuf cleanup - dtom() has now been removed, and
979 a normal malloc() is used for PCB allocations and in other places
980 where mbufs were being abused for structural allocations.
981* dup_sockaddr() now unconditionally uses M_INTWAIT instead of
982 conditionally using M_NOWAIT, making it more reliable.
983* Fix a number of USB device ref counting issues and fix issues
984 related to UMASS detaching from CAM while CAM is still active,
985 and vice-versa.
986* Optimize kern_getcwd() some to avoid a string shifting bcopy().
987* Continued work on asynch syscalls - track pending system calls
988 and make exit1() wait for them (abort support will be forthcoming).
989* **Add a negative lookup cache for NFS.** This makes a huge
990 difference for things like buildworlds where /usr/src is NFS
991 mounted, reducing (post cached) network bandwidth to 1/10 what
992 it was before.
993* Add the '-l' option to the 'resident' command, listing all
994 residented programs and their full paths (if available). -Hiten.
995* **Implement the 'rconfig' utility (see the manual page)** - for
996 automatic search/config-script downloading and execution, which
997 makes installing a new DFly box from CDBoot a whole lot easier
998 when you are in a multi-machine environment.
999* Revamp the BIO b_dev assignment and revamp the 'disk' layer.
1000 Instead of overloading the raw disk device the disk layer now
1001 creates a new device on top of the raw disk device and takes over
1002 the (user accessible) CDEV (major,minor). The disk layer does
1003 its work and reassigned b_dev to the raw device. biodone() now
1004 unconditionally setes b_dev to NODEV and all I/O ops are required
1005 to initialize b_dev prior to initiating the op. This is
1006 precursor work to our DEV layering and messaging goal.
1007* Fix the rootfs search to specify the correct unit number rather
1008 then using unit 0, because CDEVSW lookups now require a valid
1009 minor number (CDEVSW's are now registered with a minor number
1010 range and the same major number can be overloaded as long as the
1011 minor ranges do not conflict). Fix by Hiroki Sato.
1012* Fix a wiring related page table memory leak in the VM system.
1013* Fix a number of ^T related panics.
1014* Properly ref-count all devices.
1018## Sun 2 May 2004
1020* NEWTOKENs replace old tokens. The new token abstraction provides
1021 a serialization mechanism that persists across blocking conditions.
1022 This is not a lock, but more like an MP-capable SPL, in that
1023 if you block you will lose serialization but then regain it when
1024 you wakeup again. This means that newtokens can be used for
1025 serialization without having to make lower level subsystems
1026 aware of tokens held by higher level subsystems in the call
1027 stack. This represents a huge simplification over the FreeBSD-5
1028 mutex model.
1029* Support added for the Silicon Image SATA controller.
1030* DragonFly now supports 16 partitions per slice (up from 8).
1031* Wildcarded sockets have been split from the TCP/UDP connection
1032 hash table, and the listen table is now replicated across
1033 cpus for lockless access.
1034* UDP sendto() without a connect no longer needs to make a
1035 temporary connect inside the kernel, greatly improving UDP
1036 performance in this case.
1037* NFS performance has been greatly improved.
1038* Fix some major bugs in the USB/SIM code, greatly improving
1039 the reliability of USB disk keys and related devices.
1040* NEWCARD has been brought in from FreeBSD-5.
1041* A bunch of userland scheduler performance issues have been fixed.
1042* Major syscall procedural separation has been completed, separating
1043 the user interfacing portion of a syscall from the in-kernel
1044 core support, allowing the core support functions to be directly
1045 called from emulation code instead of using the stackgap junk.
1046* An optimized MMX and XMM kernel bcopy has been implemented and
1047 tested. Most of i386/i386/support.s has been rewritten and a
1048 number of FP races in that and npxdna() have been closed.
1049* Brought in the SFBUF facility from FreeBSD-5, including all of
1050 Alan Cox's (the FBsd Alan Cox) improvements, plus additional
1051 improvements for cpu-localized invlpg calls (and the ability
1052 to avoid such calls when they aren't needed).
1053* A major PIPE code revamp has occured, augmenting the SFBUF
1054 direct-copy feature with a linear map. Peak standard pipe
1055 throughput with 32KB buffers on an AMD64 3200+ now exceeds
1056 4GBytes/sec.
1057* Implement XIO, which is a multi-page buffer wrapper for SFBUFs
1058 and will eventually replace linear buffer management in the
1059 buffer cache as well as provide linear buffer mappings for other
1060 parts of the system.
1061* Added a localized cpu pmap_qenter() called pmap_qenter2() which
1062 is capable of maintaining a cpumask, to avoid unnecessary invlpg's
1063 on target-side linear maps. It is currently used by the PIPE
1064 code and the CAPS code.
1065* Joerg has brought in most of the KOBJ extensions from FreeBSD-5.
1066* Major continuing work by Jeff on the threading and partitioning
1067 of the network stack. Nearly the whole stack is now
1068 theoretically MP safe, with only a few niggling issues before we
1069 can actually start turning off the MP lock.
1070* The last major element of the LWKT messaging system,
1071 lwkt_abortmsg() support, is now in place. Also it is now possible
1072 to wait on a port with PCATCH.
1073* Hiten has revamped the TCP protocol statistics, making them
1074 per-cpu along the same lines that we have made vmstats per-cpu.
1075* propolice has been turned on by default in GCC3.
1076* CAPS (DragonFly userland IPC messaging) has been revamped to
1077 use SFBUFs.
1078* acpica5 now compiles, still needs testing.
1079* SYSTIMERS added, replacing the original hardclock(), statclock(),
1080 and other clocks, and is now used as a basis for timing in the
1081 system. SYSTIMERS provide a fine-grained, MP-capable,
1082 cpu-localizable and distributable timer interrupt abstraction.
1083* Fix RTC write-back issues that were preventing 'ntpdate' changes
1084 from being written to the RTC in certain cases.
1085* Finish most of the namecache topology work. We no longer need
1086 the v_dd and v_ddid junk. The namecache topology is almost
1087 ready for the next major step, which will be namespace locking
1088 via namecache rather then via vnode locks.
1089* Add ENOENT caching for NFS, greatly reducing the network overhead
1090 (by a factor of 5x!!!) required to support things like
1091 buildworld's using an NFS-mounted /usr/src.
1092* Jeff has added predicate messaging requests to the network
1093 subsystem, which allows us to convert situations which normally
1094 block, like connect(), to use LWKT messages.
1095* Lots of style cleanups, primarily by Chris Pressey.
1099## Sun 15 February 2004
1101* Newcard is being integrated.
1102* A longstanding bug in PCI bus assignments which affects larger
1103 servers has been fixed.
1104* The IP checksum code has been rewritten and most of it has been
1105 moved to machine-independent sections.
1106* A general machine-independent CPU synchronization and rendezvous
1107 API has been implemented. Older hardwired IPIs are slowly being
1108 moved to the new API.
1110* A new 'SysTimer' API has been built which is both MP capable
1111 and fine-grained. 8254 Timer 0 is now being used for fine-grained
1112 timing, while Timer 2 is being used for timebase. Various
1113 hardwired clock distribution code is being slowly moved to the
1114 new API. hardclock and statclock have already been moved.
1115* A long standing bug in the NTP synchronization code has been fixed.
1116 NTPD should now lock much more quickly.
1117* Clock interrupt distribution has been rewritten. Along with this,
1118 IPI messaging now has the ability to pass an interrupt frame
1119 pointer to the target function. Most of the old hardwired
1120 clock distribution code has been ripped out.
1121* nanosleep() is now a fine-grained sleep. After all, what's the
1122 point of having a nanosleep() system call which is only capable
1123 of tick granularity?
1124* Critical fixes from FreeBSD RELENG_4 integrated by Hiten Pandya.
1125* Firewire subsystem updated by a patchset from Hidetoshi Shimokawa.
1127* USB subsystem has been synced with FreeBSD 5 and NetBSD.
1128* GCC 3.3.3 and Binutils 2.14 have been integrated into base.
1129* An aggregated Client/Server Directory Services syscall API has
1130 been completed.
1131* An amiga-style 'resident' utility program + kernel support has
1132 been implemented, and prelinking support has been removed
1133 (because the resident utility is much better). You can make any
1134 dynamically loaded ELF binary resident with this utility. The
1135 system will load the program and do all shared library mappings
1136 and relocations, then will snapshot the vmspace. All future
1137 executions of the program simply make a copy of the saved
1138 vmspace and skip almost right to main(). Kernel overhead is
1139 fairly low, also. It still isn't as fast as a static binary
1140 but it is considerably faster then non-resident dynamic binaries.
1143## Mon 1 December 2003
1145A great deal of new infrastructure is starting to come to fruition.
1148* We have a new CD release framework (/usr/src/nrelease) in.
1149 Development on the new framework is proceeding. Basically the
1150 framework is based on a fully functioning live CD system
1151 and will allow us to build a new installation infrastructure
1152 that is capable of leveraging all the features available in
1153 a fully functioning system rather then being forced to use
1154 a lobotomized set.
1155* A new IPC framework to reliably handle things like password
1156 file lookups is proceeding. This framework is intended to
1157 remove the need for DLL or statically-linked PAM/NSS and at
1158 the same time make maintainance and upgrades easier and more
1159 portable.
1160* The FreeBSD-5 boot infrastructure has been ported and is now
1161 the default.
1162* RCNG from FreeBSD-5 has been ported and is now the default.
1163* Work is proceeding on bringing in ATAng from FreeBSD-4, with
1164 additional modifications required to guarantee I/O in low
1165 memory situations. That is, it isn't going to be a straight
1166 port. The original ATA code from FreeBSD-4.8, which we call
1167 ATAold, has been given interim fixes to deal with the low memory
1168 and PIO backoff issues so we don't have to rush ATAng.
1171## Sat 18 October 2003
1173Wow, October already! Good progress is being made on several fronts.
1175* K&amp;R function removal.
1176* VM function cleanups by Hiten Pandya.
1177* General kernel API cleanups by David Rhodus.
1178* Syscall Separation work by David Reese.
1180* Removal of stackgap code in the Linux Emulation by David Reese.
1181* Networking work by Jeffrey Hsu.
1182* Interrupt, Slab Allocator stabilization.
1183* Introduction of _KERNEL_STRUCTURES ... a better way for
1184 userland programs to access kernel header files rather
1185 then them setting _KERNEL.
1186* Bring the system uptodate on security issues (David Rhodus, others)
1187* NFS peformance improvements by David Rhodus and Hiten Pandya.
1189* GUPROF and kldload work in the kernel by Hiten Pandya.
1190* Major progress on the checkpointing code in the kernel
1191 primarily by Kip Macy.
1192* All work through this moment has been stabilized with major
1193 input from David Rhodus.
1196Matt's current focus continues to be on rewriting the namecache code.
1197Several intermediate commits have already been made but the big changes
1198are still ahead.
1200Galen has started experimenting with userland threads, by porting the
1201LWKT subsystem (which is mostly self contained) to userland.
1203## Wed 27 August 2003 - Slab Allocator, __P removal
1206* DragonFly now has slab allocator for the kernel! The allocator is
1207 about 1/3 the size of FreeBSD-5's slab allocator and features per-cpu
1208 isolation, mutexless operation, cache sensitivity (locality of reference),
1209 and optimized zeroing code.
1211* The core of the slab allocator is MP safe but at the moment we still
1212 use the malloc_type structure for statistics reporting which is not
1213 yet MP safe, and the backing store (KVM routines) are not MP safe.
1214 Even, so making the whole thing MP safe is not expected to be difficult.
1217* Robert Garrett has made great process removing __P(), Jeffrey has been
1218 working on the nework stack, David has been backporting bug fixes from
1219 FreeBSD and doing other cleanups, and Hiten and Jeroen have been
1220 investigating documentation and source code reorganization.
1223## Sun 10 August 2003 - Source Reorganization
1225* A major source tree reorganization has been accomplished, including
1226 separation of device drivers by functionality, moving filesystems into
1227 a vfs/ subdirectory, and the removal of the modules/ subdirectory with
1228 an intent to integrate the module makefiles into the normal sys/
1229 tree structure.
1231* Work on syscall messaging is ongoing and we will soon hopefully have
1232 a fully working demonstration of asynch messaging.
1235## 09 July 2003 to 22 July 2003 - Misc work, DEV messaging
1237* A large number of commits related to the messaging infrastructure have
1238 been made, and DEV has been message encapsulated (though it still runs
1239 synchronously).
1241* Announced the official start of work on the userland threading API:
1242 <a href="misc/uthread.txt">misc/uthread.txt</a>
1246## 27 June 2003 to 09 July 2003 - MP operation
1248This section contains work completed so far. Some items have not yet
1249been integrated into the next section up.
1251* **(done)** Get it so user processes can run simultaniously
1252 on multiple cpus. User processes are MP, the kernel is still
1253 mostly UP via the MP lock.
1254* **(done)** Run normal interrupts and software interrupts
1255 in their own threads.
1257* **(done)** Implement interrupt preemption with a
1258 block-return-to-original-thread feature (which is more
1259 like BSDI and less like 5.x).
1260* **(done)** Finish separating the userland scheduler
1261 from LWKT. The userland scheduler now schedules one LWKT
1262 thread per cpu. Additionally, the realtime, normal, and
1263 idle priority queues work as expected and do not create
1264 priority inversions in the kernel. Implement a strict
1265 priority+rr model for LWKTs and assign priorities to
1266 interrupts, software interrupts, and user processes running
1267 in kernel mode. Deal with ps, systat, and top. Fix
1268 bugs in the sysctl used to retrieve processes. Add threads
1269 to the information returned by said sysctl.
1270* Replace the UIO structure with a linked list of VM objects,
1271 offsets, and ranges, which will serve for both system and
1272 user I/O requests.
1273* Move kernel memory management into its own thread(s),
1274 then pull in the SLAB allocator from DUX3, implement a
1275 per-cpu cache, and get rid of zalloc*().
1276* Implement virtual cpus, primarily for testing purposes.
1278* (done) Separate scheduling of user processes from the
1279 LWKT scheduler.
1280 i.e. only schedule one user process in the LWKT scheduler at
1281 a time per cpu.
1282* Change over to a flat 64 bit I/O path.
1283* (done) Get real SMP working again... implement the token
1284 transfer matrix between cpus, keep the BGL for the moment but
1285 physically move the count into the thread structure so it
1286 doesn't add (much) to our switching overhead.
1287* Fix BUF/BIO and turn I/O initiation into a message-passing
1288 subsystem. Move all DEVices to their own threads and
1289 implement as message-passing. VFS needs a major overhaul
1290 before VFS devices can be moved into their own threads to
1291 the reentrant nature of VFS.
b6af6399 1294## 17 June 2003 to 26 June 2003
1296* Add light weight kernel threads to the tree.
1298This work has been completed. It involved creating a clearly defined
1299thread/process abstraction.
1301* **(done)** embed a thread structure in the proc structure.
1302* **(done)** replace the curproc global with curthread, create
1303 macros to mimic the old curproc in C code.
1304* **(done)** Add idlethread. curthread is never NULL now.
1305* **(done)** replace the npxproc global with npxthread.
1307* **(done)** Separate the thread structure from the proc structure.
1308* **(done)** remove the curpcb global. Access it via curthread.
1309 ('curthread' will be the only global that needs to be
1310 changed when switching threads. Move the PCB to the end
1311 of the thread stack.
1312* **(done)** npxproc becomes npxthread.
1313* **(done)** cleanup globaldata access.
1315* **(done)** Separate the heavy weight scheduler from the thread
1316 scheduler and make the low level switching function operate
1317 directly on threads and only threads. Heavy weight process
1318 switching (involving things like user_ret, timestamps,
1319 and so forth) will occur as an abstraction on top of the
1320 LWKT scheduler. swtch.s is almost there already.
1321 The LWKT switcher only messes with basic registers and
1322 ignores the things that are only required by full blown
1323 processes ( debug regs, FP, common_tss, and CR3 ). The heavy
1324 weight scheduler that deals with full blown process contexts
1325 handles all save/restore elements. LWKT switch times are
1326 extremely fast.
1327* **(done)** change all system cals from (proc,uap) to (uap).
1328* **(done)** change the device interface to take threads instead of
1329 procs (d_thread_t is now a thread pointer). Change the
1330 select interface to take thread arguments instead of procs
1331 (interface itself still needs to be fixed). Consolidate
1332 p_cred into p_ucred. Consolidate p_prison into p_ucred.
1333 Change suser*() to take ucreds instead of processes.
1334* **(done)** Move kernel stack management to the thread structure.
1335 Note: the kernel stack may not be swapped. Move the pcb to
1336 the top of the kernel stack and point to it from
1337 the thread via td_pcb (similar to FreeBSD-5).
1339* **(done)** Get rid of the idiotic microuptime and related
1340 crap from the critical path in the switching code.
1341 Implement statistical time statistics from the stat clock
1342 interrupt that works for both threads and processes without
1343 ruining switching performance (very necessary since there
1344 is going to be far more kernel->kernel done with this
1345 design, but it also gets rid of a serious eyesore that
1346 has been in the FreeBSD tree for a long time).
1347* **(done)** Translate most proc pointers into thread pointers
1348 in the VFS and DEV subsystems (which took all day to do).
1349* **(done)** Cleanup VFS and BUF. Remove creds from calls that should
1350 not pass them, such as close, getattr, vtruncbuf, fsync, and
1351 createvobject. Remove creds from struct buf. All of the
1352 above operations can be run in any context, passing the
1353 originator's cred makes no sense.
1354* **(95% done and tested)** Remove all use of **curproc** and
1355 **curthread** from the VFS
1356 and DEV subsystems in preparation for moving them to
1357 serializable threads. Add thread arguments as necessary.
1359* **(95% done and tested)** Finish up using thread references
1360 for all subsystems that
1361 have no business talking to a process. e.g. VM, BIO.
1362 Deal with UIO_COPY.
1363* **(done)**Make tsleep/wakeup work with pure threads.
1364* **(done, needs more testing. buildworld succeeds on test box)**
1365 Move kernel threads (which are implemented as heavy-weight
1366 processes to LWKT threads).
b6af6399 1369## 16 June 2003 -
1371* Completed repository fork, $Tag change, and various cleanup.
1373* Creating a new repository required a bit of work. First the RELENG_4
1374 tree had to be checked out. Then a new cvs repository was created
1375 with a new cvs tag. Then a combination of scripts and manual work
1376 (taking all day) was used to move the $FreeBSD tags into comments
1377 and add a new tag to all the source files to represent the new
1378 repository. I decided to cleanup a large number of FBSDID and
1379 rcsid[] declarations at the same time, basically moving all tag
1380 descriptions into comments with the intent that a linking support
1381 program would be written later to collect tags for placement in
1382 binaries. Third party tags in static declarations which contained
1383 copyright information were retained as static declarations.
1385* Some minor adjustments to the syscall generator was required, and
1386 I also decided to physically remove UUCP from the tree.
1388* Finally, buildworld and buildkernel were made to work again and
1389 the result was checked in as rev 1.2 (rev 1.1 being the original
1390 RELENG_4 code).