dragonfly.git
13 years agosystat - Add more spacing for the Load
Matthew Dillon [Sun, 15 Aug 2010 19:18:48 +0000 (12:18 -0700)]
systat - Add more spacing for the Load

* Adjust load display so three-digit loads do not crush the display

13 years agokernel - split the pbuf subsystem into two (kva and non-kva)
Matthew Dillon [Sun, 15 Aug 2010 18:57:53 +0000 (11:57 -0700)]
kernel - split the pbuf subsystem into two (kva and non-kva)

Most pbufs do not require kva reservations.  Split the API to provide
pbufs without kva reservations out of a much larger pool of pbufs, fixing
deadlock issues with the DM subsystem.

Note in particular that the hammer reblocker can eat upwards of 5000 pbufs
when operating on multi-layered DM-based storage, which blows away the
256 pbufs normally available with kva reservations that the old API had.

* Getpbuf() / trypbuf() now return pbufs without KVA reservations.

* Add getpbuf_kva() and trypbuf_kva() to get pbufs with KVA reservations.

* Fixes pbuf deadlocks in the low level I/O subsystem, particularly DM
  crypt, stripe, and mirror.

13 years agokernel - dm_crypt - Pre-zero control structures, disable B_HASBOGUS handling
Matthew Dillon [Sun, 15 Aug 2010 15:24:13 +0000 (08:24 -0700)]
kernel - dm_crypt - Pre-zero control structures, disable B_HASBOGUS handling

* Don't take any chances, pre-zero control structures and make sure the
  dmtc is completely initialized.

* Comment out the B_HASBOGUS read-decrypt bogus_page handling code.
  It doesn't work yet and it turns out that it might not even be
  necessary because the decrypter works on 512-byte sub blocks.

13 years agokernel - Add overlapping buffer test to the buffer cache
Matthew Dillon [Sun, 15 Aug 2010 15:20:37 +0000 (08:20 -0700)]
kernel - Add overlapping buffer test to the buffer cache

* Add a test for overlapping buffer in bgetvp().

* Add a sysctl vfs.check_buf_overlap which defaults to enabling the
  check invasively.  Overlapping buffer cache buffers can be deadly.

  0 - disabled
  1 - checked and reported
  2 - checked and panic (invasive) (default)

13 years agoHAMMER VFS - Fix over-enthusiastic cluster read
Matthew Dillon [Sun, 15 Aug 2010 15:14:53 +0000 (08:14 -0700)]
HAMMER VFS - Fix over-enthusiastic cluster read

* The block device I/O was over-enthusiastic in calling cluster_read()
  and could wind up creating the buffers of the wrong size which
  would then overlap the address space later buffer requests for the
  right size.

  This could result in the corruption of large-data (64K) blocks,
  usually causing a hammer reblock to fail with a CRC error but
  not corrupting the actual filesystem on-media.

  Meta data could not usually get corrupted by this unless the
  cluster-read happened to cross a large-block (8MB) boundary.

* Particularly easy to reproduce with the dm_crypt module due to
  crypt overheads.

* Fixed by disallowing read-aheads in the large-data zone (the only
  zone which can contain a mix of 16K and 64K blocks), and ensuring
  that any other cluster_read does not cross a large-block boundary.

13 years agolvm - don't forget to include dlfcn.h
Alex Hornung [Sun, 15 Aug 2010 12:24:24 +0000 (13:24 +0100)]
lvm - don't forget to include dlfcn.h

* dlfcn is needed for some of the functions used in toolcontext.c, so
  include it on DragonFly, too.

Reported-by: Sascha Wildner (swildner@)
13 years agolvm - adjust visibility for devname
Alex Hornung [Sun, 15 Aug 2010 12:20:33 +0000 (13:20 +0100)]
lvm - adjust visibility for devname

* adjust visibility in lib/dragonfly/dev.c to show devname(), avoiding
  truncation of higher 32 bits on x86_64, since otherwise the return
  value defaults to int.

Reported-by: Sascha Wildner (swildner@)
13 years agolinuxulator - Acquire MPLOCK for ksignal
Alex Hornung [Sun, 15 Aug 2010 10:09:26 +0000 (11:09 +0100)]
linuxulator - Acquire MPLOCK for ksignal

* Don't forget to acquire mplock for ksignal, or otherwise we'll get a
  panic (MP_LOCK_HELD: Not held thread foo).

13 years agokernel - dm - Reorganize the crypt code and implement B_HASBOGUS
Matthew Dillon [Sun, 15 Aug 2010 00:51:42 +0000 (17:51 -0700)]
kernel - dm - Reorganize the crypt code and implement B_HASBOGUS

* Implement B_HASBOGUS data copying for read decryption.  When a buffer
  contains bogus pages we may or may not be able to safely decrypt data
  in-place, add some code to deal with it.

* Reorganize the crypt code.  Separate the read and write paths and
  organize the procedures in the same order as the execution stages.
  Also move some of the buffer manipulation around a little to make
  it more clear.

Submitted-by: alexh (B_HASBOGUS handling)
13 years agokernel - dm - Fix panic in dm_target_crypt
Matthew Dillon [Sat, 14 Aug 2010 23:19:17 +0000 (16:19 -0700)]
kernel - dm - Fix panic in dm_target_crypt

* Fix a misordered kfree()

* Minor cleanup

Reported-by: alexh
13 years agocryptsetup - fix buffer overflow
Alex Hornung [Sat, 14 Aug 2010 22:01:47 +0000 (23:01 +0100)]
cryptsetup - fix buffer overflow

* fix a buffer overflow introduced during the porting. Linux originally
  did digest = gcry_md_read(hd, hash_id); memcpy(dst, digest, len);
  I ported that to do EVP_DigestFinal directly into dst, causing
  corruption. Avoid this by writing to some intermediate (large) buffer
  and then using memcpy like Linux does.

13 years agocryptsetup - fix issue with uuid
Alex Hornung [Sat, 14 Aug 2010 20:17:27 +0000 (21:17 +0100)]
cryptsetup - fix issue with uuid

* the whole codepath related to uuid was a bit messy. Nothing should've
  been done if uuid==NULL and the result was passed in as integer, not
  pointer...

13 years agokernel - netif - temporary hack for IFT_ETHER
Matthew Dillon [Sat, 14 Aug 2010 19:10:39 +0000 (12:10 -0700)]
kernel - netif - temporary hack for IFT_ETHER

* Temporarily hack if_alloc() to allocate a struct arpcom instead of a
  struct ifnet when IFT_ETHER is passed in, until we can adjust all the
  drivers to use the ifnet->if_l2com API.

Reported-by: Johannes Hofmann <johannes.hofmann@gmx.de>
13 years agokernel - dm - Add support for BUF_CMD_FLUSH
Matthew Dillon [Sat, 14 Aug 2010 19:04:52 +0000 (12:04 -0700)]
kernel - dm - Add support for BUF_CMD_FLUSH

* Change the nestiobuf_*() code to count the number of sub-bios issued
  instead of counting the resid.  This allows the nestiobuf_*() API to
  be used to BUF_CMD_FLUSH.

* Change the nestiobuf_*() API.  The nestiobuf_start() command is now
  nestiobuf_add().  Add a nestiobuf_init(mbio) and nestiobuf_start(mbio)
  wrapper around the adds.

* Explicitly check for allowed BUF_CMD_* commands in device-mapper.c

* Add a bypass to device-mapper.c to cause certain commands to be sent
  to all tbl's (BUF_CMD_FLUSH).

* Add explicit suppor for BUF_CMD_FLUSH to dm_target_stripe.c

* Change b_resid loading semantics to load upon completion of the
  I/O.

13 years agokernel - flag when bogus_page's are present
Matthew Dillon [Sat, 14 Aug 2010 17:53:43 +0000 (10:53 -0700)]
kernel - flag when bogus_page's are present

* Flag a buffer when bogus pages are present.  Add B_HASBOGUS to b_flags.

  Certain in-place read operations, such as decryption, must allocate a
  copy if the original buffer contains bogus pages.

13 years agokernel - dm - rewrite dm_target_stripe.c to add functionality and fix bugs
Matthew Dillon [Sat, 14 Aug 2010 17:06:21 +0000 (10:06 -0700)]
kernel - dm - rewrite dm_target_stripe.c to add functionality and fix bugs

* The stripe code now allows up to 32 devices, instead of just 2
* The stripe code now properly cleans up partial configurations which
  fail in the device loop.

* Fix improper use of M_NOWAIT
* Fix improper use of malloc pool in *_status functions
* Fix improper use of namespace (atoi -> atoi64)
* Cleanup some M_DM malloc declarations

13 years agokernel - devfs - Fix dangling lock created by recent commit
Matthew Dillon [Sat, 14 Aug 2010 15:45:50 +0000 (08:45 -0700)]
kernel - devfs - Fix dangling lock created by recent commit

* devfs_freep() is not always called with devfs_lock held

Reported-by: alexh
13 years agolibdm/lvm - Fix bugs
Alex Hornung [Sat, 14 Aug 2010 08:37:54 +0000 (09:37 +0100)]
libdm/lvm - Fix bugs

* Allow adding symlinks to lvm's dev-cache by translating the symlink
  for udev/libdevattr.

* Don't include kdev_t.h in libdm since it has the wrong macros for
  MAJOR and MINOR; instead include netbsd's version.

Reported-by: Matt Dillon (dillon@)
13 years agoMerge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Sat, 14 Aug 2010 01:01:13 +0000 (18:01 -0700)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly

13 years agokernel - kqueue - fix EVFILT_WRITE notification on pipes
Matthew Dillon [Sat, 14 Aug 2010 00:59:47 +0000 (17:59 -0700)]
kernel - kqueue - fix EVFILT_WRITE notification on pipes

* The pipe code was running the wrong kqueue for write buffer
  availability notifications, so things like scp transfers were timing
  out once the write socket filled up when they tried to poll() for
  more space.

Reported-by: swildner
13 years agoFix a printf issue in bthcid(8) (fixes x86_64 buildworld).
Sascha Wildner [Fri, 13 Aug 2010 23:21:42 +0000 (01:21 +0200)]
Fix a printf issue in bthcid(8) (fixes x86_64 buildworld).

13 years agokernel - make POLLHUP conform better to OpenGroup
Matthew Dillon [Fri, 13 Aug 2010 22:06:12 +0000 (15:06 -0700)]
kernel - make POLLHUP conform better to OpenGroup

* The poll() backend now only evaluates EV_EOF for EVFILT_WRITE, and
  generates a POLLHUP which is properly mutually exclusive with
  POLLWRNORM and POLLOUT.

  We do this even for a half-closed (write-side-closed) socket.

* We no longer set POLLHUP for POLLRD* or POLLIN, instead relying on
  the userland doing the read() and checking the 0 result (which
  userland has always done historically).

  EV_EOF for EVFILT_READ cannot be used to set POLLHUP for POLLRD* or
  POLLIN because it could indicate a half-closed connection
  (read-side-closed) where writing is still allowed, which is a more
  common situation on sockets.

  We would have to add another EV_ flag to set POLLHUP for POLLRD* or
  POLLIN to detect the fully disconnected state.  The OpenGroup standard
  does allow POLLHUP to be mixed with POLLRD* or POLLIN but for now we
  just don't set it at all in that case.

  Only a POLLOUT/POLLWRNORM flag request can cause POLLHUP to be set
  for now.

13 years agokernel - More kqueue work
Matthew Dillon [Fri, 13 Aug 2010 21:05:45 +0000 (14:05 -0700)]
kernel - More kqueue work

* Only set EV_EOF in the read filter after all pending data has
  been exhausted.

* This also fixes a bug where the read filter was not setting data
  ready on EOF when data was pending in the buffer.

* Fix bugs in the poll copyout handler.  An EOF condition does not
  prevent other flags from getting set.

13 years agokernel - Some minor swap allocator / blist work
Matthew Dillon [Fri, 13 Aug 2010 21:00:06 +0000 (14:00 -0700)]
kernel - Some minor swap allocator / blist work

* Fix a bug in the 'r' option for the userland compile of the blist code

* Finish converting daddr_t's to swblk_t's in the swap code.

* Move SWAPBLK_* out of vm/vm_page.h (taken from FreeBSD) to reduce
  the number of #includes needed.  Move them to sys/blist.h

Reported-by: Ilya Dryomov <idryomov@gmail.com>
13 years agostress2 - Use statvfs to fix calculations for large filesystems
Matthew Dillon [Fri, 13 Aug 2010 15:50:03 +0000 (08:50 -0700)]
stress2 - Use statvfs to fix calculations for large filesystems

* Use statvfs to avoid available space overflow in calculation.

13 years agolibdevattr - Minimal error checking in devattr_test
Antonio Huete Jimenez [Fri, 13 Aug 2010 11:49:45 +0000 (13:49 +0200)]
libdevattr - Minimal error checking in devattr_test

13 years agolibevent - Remove from tree
Samuel J. Greear [Thu, 12 Aug 2010 15:09:37 +0000 (15:09 +0000)]
libevent - Remove from tree

* There are no longer any in-tree consumers

* Maintenance burden is moved to pkgsrc

13 years agoftp-proxy - Port from libevent to kqueue(2)/kevent(2)
Samuel J. Greear [Fri, 13 Aug 2010 01:06:38 +0000 (01:06 +0000)]
ftp-proxy - Port from libevent to kqueue(2)/kevent(2)

* Was using libevent's bufferevent* API, we let the socket buffers do
  most of the work.

13 years agobthcid - Port from libevent to kqueue(2)/kevent(2)
Samuel J. Greear [Thu, 12 Aug 2010 14:48:31 +0000 (14:48 +0000)]
bthcid - Port from libevent to kqueue(2)/kevent(2)

13 years agokernel - Fix reboot races
Matthew Dillon [Fri, 13 Aug 2010 05:45:25 +0000 (22:45 -0700)]
kernel - Fix reboot races

* Fix a pty revoke race null pointer indirection during reboot

* Fix a deadlock with devfs (introduced by a recent commit) by
  not having devfs_config() do anything if it is called from
  the message core itself.

13 years agoMerge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Fri, 13 Aug 2010 04:50:49 +0000 (21:50 -0700)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly

13 years agokernel - devfs - Fix deadlocks
Matthew Dillon [Fri, 13 Aug 2010 04:49:03 +0000 (21:49 -0700)]
kernel - devfs - Fix deadlocks

* Fix a deadlocks when the core thread attempts to free a node.  The
  devfs lock must be released when obtaining the vnode lock.  Fixes
  issues with ls -la /dev/pts when running stress2{pts}.

* Adjust devfs_config() to not attempt to message the devfs core thread
  if the core thread has terminated (fixes issues with reboot).

13 years agokernel - Add sysref assertions
Matthew Dillon [Fri, 13 Aug 2010 04:48:45 +0000 (21:48 -0700)]
kernel - Add sysref assertions

* Add more checks in the sysref code to detect use-after-free situations.

13 years agoAdjust the docs for the renaming of devfs_clone_bitmap_rst() to _put().
Sascha Wildner [Fri, 13 Aug 2010 04:33:02 +0000 (06:33 +0200)]
Adjust the docs for the renaming of devfs_clone_bitmap_rst() to _put().

13 years agokernel - Add sysref assertions
Matthew Dillon [Fri, 13 Aug 2010 04:11:20 +0000 (21:11 -0700)]
kernel - Add sysref assertions

* Add checks in the sysref code to detect use-after-free situations.

13 years agokernel - More tty and devfs race fixes
Matthew Dillon [Fri, 13 Aug 2010 04:09:13 +0000 (21:09 -0700)]
kernel - More tty and devfs race fixes

* Refactor the unix98 pty code to consolidate termination handling and
  add flags to deal with close() races.

* Rename devfs_clone_bitmap_rst() to devfs_clone_bitmap_put()

* devfs_clone_bitmap_put() now syncs devfs messages before clearing
  the bitmap bit, to ensure that related destroy_dev()s complete
  before the unit number becomes available for a new allocation
  again.

Reported-by: swildner
13 years agoHAMMER VFS - Adjust signedness of a media field for future de-dup
Matthew Dillon [Fri, 13 Aug 2010 02:02:34 +0000 (19:02 -0700)]
HAMMER VFS - Adjust signedness of a media field for future de-dup

* Make bytes_free in hammer_blockmap_layer2 a signed field and remove
  KKASSERT() that disallowed a negative value.

  This field can go negative in a future data de-dup implementation.

13 years agokernel - Fix pty registration, memory leak
Matthew Dillon [Fri, 13 Aug 2010 01:55:20 +0000 (18:55 -0700)]
kernel - Fix pty registration, memory leak

* Properly unregister the pt_ioctl before freeing it.

* Free the pt_ioctl structure after destroying the device, fixing
  a memory leak.

* Fix races in the sysctl_proc which iterates the tty list by using
  a marker.

13 years agokernel - Fix pty clone vnode leak
Matthew Dillon [Fri, 13 Aug 2010 01:20:41 +0000 (18:20 -0700)]
kernel - Fix pty clone vnode leak

* Any cloned devfs device (vn, tap, ptmx primarily) was leaving an extra
  vref on the related vnode, preventing the vnode from being recycled.

  This was due to devfs called vop_stdopen() but forgetting that
  vop_stdopen() will load the file pointer (fp), which devfs was also
  loading.  All devfs really needs to do is to replace fp->f_ops.

Reported-by: swildner
13 years agoAdd DSCHED_FQ to GENERIC.
Venkatesh Srinivas [Fri, 13 Aug 2010 00:12:49 +0000 (17:12 -0700)]
Add DSCHED_FQ to GENERIC.

13 years agokernel - Adjust krate infrastructure
Matthew Dillon [Wed, 11 Aug 2010 20:59:36 +0000 (13:59 -0700)]
kernel - Adjust krate infrastructure

* Move struct krate from sys/time.h to sys/resourcevar.h

* Remove krate from struct uidinfo, it created too much breakage
  in the buildworld.  Declare it as a static instead.

13 years agobuildworld - Fix breakage
Matthew Dillon [Wed, 11 Aug 2010 20:07:29 +0000 (13:07 -0700)]
buildworld - Fix breakage

* More cases where sys/user.h is not included early enough.

Reported-by: swildner
13 years agobuildworld - Fix breakage
Matthew Dillon [Wed, 11 Aug 2010 20:00:29 +0000 (13:00 -0700)]
buildworld - Fix breakage

* Fix some _KERNEL_STRUCTURES breakage.  This isn't a fantastic solution
  but it works.

Reported-by: swildner
13 years agoobjcache.9: Oops, .Vt actually.
Sascha Wildner [Wed, 11 Aug 2010 19:56:01 +0000 (21:56 +0200)]
objcache.9: Oops, .Vt actually.

13 years agoobjcache.9: Use .Ft for a type.
Sascha Wildner [Wed, 11 Aug 2010 19:50:49 +0000 (21:50 +0200)]
objcache.9: Use .Ft for a type.

13 years agoMerge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Tue, 10 Aug 2010 23:35:00 +0000 (16:35 -0700)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly

13 years agokernel - Add per-user file descriptor limit
Matthew Dillon [Tue, 10 Aug 2010 23:27:27 +0000 (16:27 -0700)]
kernel - Add per-user file descriptor limit

* Add kern.maxfilesperuser and kern.minfilesperproc to complement the
  kern.maxfilesperproc which already exists.  Print a warning on the
  console if a user hits the limit (rate limited).

* Track per-user files via the uidinfo structure.  Each file pointer counts
  as one file.  dup()'d and fork()'d file descriptors do not count.

* Adjust the default user limits to approximately 1/4 the system maximums
  instead of 9/10 the system maximums.  This reduces the maximum descriptors
  per process and the maximum processes per uid.  They can be raised again
  via sysctl.

* Set minfilesperproc to 8 by default.  This is a safety which guarantees
  that a process can always have at least that many open descriptors
  without tripping over kern.maxfilesperuser.

Reported-by: swildner
13 years agolibthread_xu: Convert 0 to NULL for pointer assignment.
Venkatesh Srinivas [Tue, 10 Aug 2010 22:01:01 +0000 (15:01 -0700)]
libthread_xu: Convert 0 to NULL for pointer assignment.

13 years agokernel - Fix lockf panic
Matthew Dillon [Tue, 10 Aug 2010 20:09:10 +0000 (13:09 -0700)]
kernel - Fix lockf panic

* Fix a case where the F_UNLCK call made by closef()/fdrop() could fail
  on a resource limit, preventing the locks from being properly cleared.
  This resulted in an assertion panic at exit1().

* Minor cleanups.

Reported-by: swildner
13 years agoSome cleanup in the pf manual pages.
Sascha Wildner [Tue, 10 Aug 2010 19:22:05 +0000 (21:22 +0200)]
Some cleanup in the pf manual pages.

13 years agokqueue - Fix junk kfree() in doselect()
Matthew Dillon [Tue, 10 Aug 2010 16:58:39 +0000 (09:58 -0700)]
kqueue - Fix junk kfree() in doselect()

* Fix uninitialized fields which could cause kfree() to be called on
  junk data when a copyin error occurs.

* Minor cleanups.

Reported-by: swildner
13 years agoRemove old ftp-proxy per 'make upgrade'.
Sascha Wildner [Tue, 10 Aug 2010 10:12:52 +0000 (12:12 +0200)]
Remove old ftp-proxy per 'make upgrade'.

13 years agopf(4): Remove a ', 1' that had no effect (OpenBSD dev too tired, I guess).
Sascha Wildner [Tue, 10 Aug 2010 03:20:49 +0000 (05:20 +0200)]
pf(4): Remove a ', 1' that had no effect (OpenBSD dev too tired, I guess).

13 years agoaicasm: Correct logic bug in the undefined register bit access detection code.
Sascha Wildner [Tue, 10 Aug 2010 01:45:18 +0000 (03:45 +0200)]
aicasm: Correct logic bug in the undefined register bit access detection code.

Quoting FreeBSD's commit msg:

The code in question verifies that all register write operations only change
bits that are defined (in the register definition file) for that effected
register.  The bug effectively disabled this checking.

o Fix the check by testing the opcode against all supported read ("and" based)
  operands.

o Add missing bit definitions to the aic7xxx and aic79xx register definition
  files so that the warning (treated as a fatal error) does not spuriously
  fire.

Taken-from: FreeBSD

[While I'm here, also add a missing free() and raise WARNS to 6.]

13 years agoEliminate XIG server workaround in fp_mmap; just an artifact of copying from kern_mmap.
Venkatesh Srinivas [Tue, 10 Aug 2010 00:47:49 +0000 (17:47 -0700)]
Eliminate XIG server workaround in fp_mmap; just an artifact of copying from kern_mmap.

13 years agoDocument sysctls vm.fast_fault, vm.dreadful_invltlb, lwkt.preempt_hit/_miss,
Venkatesh Srinivas [Tue, 10 Aug 2010 00:12:05 +0000 (17:12 -0700)]
Document sysctls vm.fast_fault, vm.dreadful_invltlb, lwkt.preempt_hit/_miss,
vm.swap_idle_threshold1/2.

13 years agoVM - Enable idlezero by default.
Venkatesh Srinivas [Mon, 9 Aug 2010 23:55:01 +0000 (16:55 -0700)]
VM - Enable idlezero by default.

13 years agopf.conf.5: Minor adjustments
Thomas Nikolajsen [Mon, 9 Aug 2010 22:58:03 +0000 (00:58 +0200)]
pf.conf.5: Minor adjustments

 * delete duplicate/extra lines
 * interface group not supperted on our ifconfig(8)
 * syntax adjustments

13 years agoUndo part of the last change until I understand why it broke buildkernel.
Sascha Wildner [Mon, 9 Aug 2010 20:38:39 +0000 (22:38 +0200)]
Undo part of the last change until I understand why it broke buildkernel.

13 years agoFix some mis-usage of && and || in the kernel.
Sascha Wildner [Mon, 9 Aug 2010 20:28:26 +0000 (22:28 +0200)]
Fix some mis-usage of && and || in the kernel.

13 years agosound(4): Fix typos: illeagal -> illegal
Sascha Wildner [Mon, 9 Aug 2010 20:22:45 +0000 (22:22 +0200)]
sound(4): Fix typos: illeagal -> illegal

13 years agowlan(4): Replace GNU old-style field designators with proper C.
Sascha Wildner [Mon, 9 Aug 2010 18:58:11 +0000 (20:58 +0200)]
wlan(4): Replace GNU old-style field designators with proper C.

This GCC extension has been considered obsolete since 1993.

13 years agokernel - Move KNOTE() out of interrupt path
Samuel J. Greear [Mon, 9 Aug 2010 18:00:33 +0000 (18:00 +0000)]
kernel - Move KNOTE() out of interrupt path

13 years agopf: Fix if_pfsync to compile
Jan Lentfer [Mon, 9 Aug 2010 14:02:09 +0000 (16:02 +0200)]
pf: Fix if_pfsync to compile

if_pfsync has to be enabled in the
kernel config file.
It passed the last update unattended.
This patch let's it compile, but function
is untested, also tdb functioniality
has been removed.

13 years agopf(4): Add missing file to sys/conf/files for building pf into the kernel.
Sascha Wildner [Mon, 9 Aug 2010 08:04:48 +0000 (10:04 +0200)]
pf(4): Add missing file to sys/conf/files for building pf into the kernel.

13 years agopfctl(8): Fix some little printf() issues on x86_64 (fixes buildworld).
Sascha Wildner [Sun, 8 Aug 2010 21:24:30 +0000 (23:24 +0200)]
pfctl(8): Fix some little printf() issues on x86_64 (fixes buildworld).

13 years ago.gitignore: Add *.bak, *.core, *.old and .depend.
Sascha Wildner [Sun, 8 Aug 2010 20:39:12 +0000 (22:39 +0200)]
.gitignore: Add *.bak, *.core, *.old and .depend.

13 years ago.gitignore: Sort alphabetically.
Sascha Wildner [Sun, 8 Aug 2010 20:38:18 +0000 (22:38 +0200)]
.gitignore: Sort alphabetically.

13 years agoftp-proxy(8): Remove leftover .depend.
Sascha Wildner [Sun, 8 Aug 2010 20:29:36 +0000 (22:29 +0200)]
ftp-proxy(8): Remove leftover .depend.

13 years agoftp-proxy: Update to OpenBSD 4.1
Jan Lentfer [Sun, 8 Aug 2010 16:47:04 +0000 (18:47 +0200)]
ftp-proxy: Update to OpenBSD 4.1

ftp-proxy has been moved from libexec/ to usr.sbin/

13 years agopflogd: Update to OpenBSD 4.1
Jan Lentfer [Sun, 8 Aug 2010 16:43:17 +0000 (18:43 +0200)]
pflogd: Update to OpenBSD 4.1

13 years agokernel - Fix kqueue panic on signal event knote
Matthew Dillon [Sun, 8 Aug 2010 19:43:04 +0000 (12:43 -0700)]
kernel - Fix kqueue panic on signal event knote

* sig_filtops filters on a process, not a file descriptor.  Remove
  the FILTEROP_ISFD flag.

  This should fix a panic when the kqueue is closed while the signal
  event is still active.

Reported-by: Jan Lentfer <Jan.Lentfer@web.de>
13 years agoMerge branch 'master' of git://island.quantumachine.net/dragonfly
Jan Lentfer [Sun, 8 Aug 2010 19:28:24 +0000 (21:28 +0200)]
Merge branch 'master' of git://island.quantumachine.net/dragonfly

13 years agoFix a panic in the varsym_get(2) syscall.
Sascha Wildner [Sun, 8 Aug 2010 18:13:07 +0000 (20:13 +0200)]
Fix a panic in the varsym_get(2) syscall.

In varsymfind(), release the lock only if vss != NULL. This could be
the case if garbage was passed in the mask.

13 years agoloader - Use PTOV() to convert physical addresses to virtual addresses.
Antonio Huete Jimenez [Sat, 7 Aug 2010 19:37:37 +0000 (21:37 +0200)]
loader - Use PTOV() to convert physical addresses to virtual addresses.

Dragonfly-bug: <http://bugs.dragonflybsd.org/issue1683>
Submitted-by: Sascha Wildner <swildner@>
Obtained-from: FreeBSD

13 years agoFix a panic in the __getcwd(2) syscall.
Sascha Wildner [Sun, 8 Aug 2010 09:19:09 +0000 (11:19 +0200)]
Fix a panic in the __getcwd(2) syscall.

buflen has to be unsigned so that large values do not become negative
and sneak past the MAXPATHLEN check.

13 years agoFix panics in two syscalls, caps_sys_get(2) and caps_sys_wait(2).
Sascha Wildner [Sun, 8 Aug 2010 08:43:08 +0000 (10:43 +0200)]
Fix panics in two syscalls, caps_sys_get(2) and caps_sys_wait(2).

Only call caps_drop() if caps != NULL.

13 years agopf: Update packet filter to the version that comes with OpenBSD 4.1
Jan Lentfer [Sun, 8 Aug 2010 07:44:38 +0000 (09:44 +0200)]
pf: Update packet filter to the version that comes with OpenBSD 4.1

    The original OpenBSD 4.1 defaults to "keep state flags S/SA" for
    all pass rules. In contrast to that we default to "no state". As
    in earlier verions of pf in DragonFly the default keep-state
    policy can still be set with the keep-policy option (e.g. "set
    keep-policy keep state (pickups)").

    DragonFly additions to pf have been kept: fairq support,
    pickups.

Detailed Info on changes/additions:
* ALTQ: Fix altq to work with pf_mtag
Patch by Matthew Dillon
* libkern: Revert commit e104539
strchr was added to libkern.h together with strrch
* net/if.h: add interface groups
Imported from FreeBSD.
* netinet6/in6.h: add macros
IN6_IS_ADDR_MC_INTFACELOCAL
IN6_IS_SCOPE_EMBED
PV6_ADDR_SCOPE_INTFACELOCAL
* sys/libkern.h: Add strchr and strrchr as inline functions
Brought in from FreeBSD
* sys/net/if_var.h: Import interface groups
Import interface groups and event handlers from FreeBSD
* sys/net/if_var.h: add if_pf_kif, if_groups to struct ifnet
obtained from: Open/FreeBSD
* net/if_types.h: add IFT_ENC to non-IATA-assignments
obtained from Open/FreeBSD
* net/bpf.c: add bpf_mtap_hdr from OpenBSD
Con up a minimal dummy header to pacify bpf.  Allocate
(only) a struct m_hdr on the stack.

13 years agore(4): fix re_setmulti(): PCIe mcast hash registers were swapped
Thomas Nikolajsen [Sun, 8 Aug 2010 07:41:05 +0000 (09:41 +0200)]
re(4): fix re_setmulti(): PCIe mcast hash registers were swapped

Dragonfly-bug: <http://bugs.dragonflybsd.org/issue1799>
Submitted-by: "Mitja Horvat <pinkfluid@>"
Obtained-from: FreeBSD

13 years agodaemon(8): add missing break
Thomas Nikolajsen [Sun, 8 Aug 2010 06:45:56 +0000 (08:45 +0200)]
daemon(8): add missing break

DragonFly-bug: <http://bugs.dragonflybsd.org/issue1800>
Submitted-by: "Mitja Horvat <pinkfluid@>"
13 years agoiwn.4: Miscellaneous adjustments.
Sascha Wildner [Sat, 7 Aug 2010 03:14:29 +0000 (05:14 +0200)]
iwn.4: Miscellaneous adjustments.

* Add some words about iwn6050fw.

* Comment out information about building the firmwares separately into
  the kernel. We don't support that (yet).

13 years agoiwn(4): Add Makefile for building iwn6050fw.ko.
Sascha Wildner [Sat, 7 Aug 2010 03:12:38 +0000 (05:12 +0200)]
iwn(4): Add Makefile for building iwn6050fw.ko.

13 years agoAdd iwn(4) and iwnfw(4) to LINT.
Sascha Wildner [Sat, 7 Aug 2010 03:11:48 +0000 (05:11 +0200)]
Add iwn(4) and iwnfw(4) to LINT.

13 years agoconfig - Fix typo in iwn6050 clean target.
Joe Talbott [Sat, 7 Aug 2010 02:21:23 +0000 (22:21 -0400)]
config - Fix typo in iwn6050 clean target.

13 years agoUnbreak buildworld.
Sascha Wildner [Fri, 6 Aug 2010 23:26:40 +0000 (01:26 +0200)]
Unbreak buildworld.

wlan_token shall be visible to the kernel only.

13 years agoiwn - Fix warning related to token work.
Joe Talbott [Sun, 25 Jul 2010 22:50:05 +0000 (18:50 -0400)]
iwn - Fix warning related to token work.

13 years agoiwn - Cleanup sysctl tree on detach.
Joe Talbott [Sun, 25 Jul 2010 22:49:35 +0000 (18:49 -0400)]
iwn - Cleanup sysctl tree on detach.

13 years agoiwn - Clean up memory freeing.
Joe Talbott [Sun, 25 Jul 2010 22:27:29 +0000 (18:27 -0400)]
iwn - Clean up memory freeing.

* Mark variables with NULL after freeing.
* Properly free dma memory.

13 years agoiwn - Convert to use the new wlan layer's global token.
Joe Talbott [Sun, 25 Jul 2010 20:09:21 +0000 (16:09 -0400)]
iwn - Convert to use the new wlan layer's global token.

13 years agowlan - Convert lockmgr locking to a global token.
Joe Talbott [Sun, 25 Jul 2010 20:08:19 +0000 (16:08 -0400)]
wlan - Convert lockmgr locking to a global token.

Use the lwkt_token API to synchronize the wlan layer entry points.

Discussed-With: dillon

13 years agoiwn - Bring in recent changes from FreeBSD.
Joe Talbott [Sun, 25 Jul 2010 19:29:52 +0000 (15:29 -0400)]
iwn - Bring in recent changes from FreeBSD.

Taken-From: FreeBSD

13 years agoiwn - Lock callout iwn_timer_timeout().
Joe Talbott [Sat, 24 Jul 2010 14:48:09 +0000 (10:48 -0400)]
iwn - Lock callout iwn_timer_timeout().

13 years agoiwn - Update firmware for 6000 chipset.
Joe Talbott [Tue, 20 Jul 2010 13:38:30 +0000 (09:38 -0400)]
iwn - Update firmware for 6000 chipset.

13 years agoiwnfw - Update 6000 firmware and add 6050 firmware
Joe Talbott [Tue, 20 Jul 2010 12:43:36 +0000 (08:43 -0400)]
iwnfw - Update 6000 firmware and add 6050 firmware

13 years agoAdd some more *at() system calls.
Sascha Wildner [Fri, 6 Aug 2010 14:44:50 +0000 (16:44 +0200)]
Add some more *at() system calls.

mkdirat(2), mkfifoat(2), mknodat(2), readlinkat(2), symlinkat(2)

Missing still (at least): linkat(2)

Documentation-changes-from: FreeBSD

13 years agoRemove <sys/selinfo.h> per 'make upgrade'.
Sascha Wildner [Fri, 6 Aug 2010 14:01:55 +0000 (16:01 +0200)]
Remove <sys/selinfo.h> per 'make upgrade'.

13 years agokernel - Clean up pipe filters (fixes last)
Samuel J. Greear [Fri, 6 Aug 2010 12:29:24 +0000 (12:29 +0000)]
kernel - Clean up pipe filters (fixes last)

* Remove extra pipe space calculation

* Move disconnected pipe check before token acquisition

13 years agokern - Clean up pipe filters
Samuel J. Greear [Fri, 6 Aug 2010 12:16:22 +0000 (12:16 +0000)]
kern - Clean up pipe filters

* Do not check ki_note, just issue the KNOTE and let kq sort it out.

* Remove an unmatched rel_mplock().

* Take the r and w tokens for the pipe in question before poking around for
  states and sizes.

* Reverse the pipe end that the write filter assumes it will receive, this
  matches the semantics of the previous .fo_poll implementation.

13 years agokernel - Silently swallow EOPNOTSUPP filter errors for select(2) in all cases
Samuel J. Greear [Fri, 6 Aug 2010 09:19:38 +0000 (09:19 +0000)]
kernel - Silently swallow EOPNOTSUPP filter errors for select(2) in all cases

* This fixes hald

Reported-by: Rumko
13 years agokernel - make poll return only explicitly asked for events
Samuel J. Greear [Fri, 6 Aug 2010 09:15:15 +0000 (09:15 +0000)]
kernel - make poll return only explicitly asked for events

* Previously we were returning all events that were analogous to each other,
  now we return only what was explicitly asked for.

* This fixes pulseaudio

Reported-by: Rumko