Thomas Nikolajsen [Wed, 11 Feb 2009 22:54:08 +0000 (23:54 +0100)]
Makefile.usr: add path to pkgsrc-all tar file
Thomas Nikolajsen [Wed, 11 Feb 2009 22:02:08 +0000 (23:02 +0100)]
nrelease: clean up Makefile and sync man pages & Makefile.usr
* nrelease/Makefile: fix a few typos,
change name of `syssrcs` target to `srcs`, to reflect current use and
remove traces of old installer_* targets.
* etc/Makefile: update name of src-sys tar file from nrelease and
add targets for newly added src-all and pkgsrc-all tar files.
* release.7: update to current method: only binary package downloaded is cvsup
and add description of a few more variables.
* development.7: add short description of /usr/Makefile.
Matthew Dillon [Wed, 11 Feb 2009 20:15:16 +0000 (12:15 -0800)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Wed, 11 Feb 2009 20:12:34 +0000 (12:12 -0800)]
Change pkg_radd to postfix DragonFly-x.x.x onto the path. This will work in
conjunction with the new pkgbox redirect CGI which no longer post-pends
the OS path, and softlinks in the packages/ sub-directory to point to the
correct quarterly release.
Sascha Wildner [Wed, 11 Feb 2009 20:09:39 +0000 (21:09 +0100)]
vnconfig: Print better msg when neither regular file nor -S is specified.
If no regular file is specified, vn(4) by default will use swap for backing
store. If swap size is missing also, print a smart warning instead of the
cryptic: "Numerical argument out of domain".
Submitted-by: Stathis Kamperis <ekamperi@gmail.com>
Dragonfly-bug: <http://bugs.dragonflybsd.org/issue1279>
Simon Schubert [Wed, 11 Feb 2009 10:25:45 +0000 (11:25 +0100)]
hammer boot2: reset file offset each time we're looking up a new inode
The boot2/ufs code used the implicit assumption that the file offset
gets reset each time a new inode is looked up. We have to follow this
behavior, or else reads will always start from the last offset of the
last file.
Matthew Dillon [Wed, 11 Feb 2009 10:02:08 +0000 (02:02 -0800)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Wed, 11 Feb 2009 09:53:52 +0000 (01:53 -0800)]
Fix long-standing bug in boot1 code - can read junk from fake partition table
boot1 reads the MBR to locate the BSD partition type (0xA5). However, to
reduce the size of the boot1 code the 32 bit LBA for the MBR was being
loaded via the fake partition table at label 'part4', which was assumed to
contain a LBA of 0. Unfortunately this portion of the boot1 code is not
usually written by the disklabel program and may contain garbage.
For the last few years we have worked around the issue by zeroing out
the label area before installing a new label. We still have to do this
to avoid disklabel32/disklabel64 confusion, but with this fix forgetting
to zero the area should not cause a properly installed disklabel to fail
to boot properly.
To fix the problem, add the necessary instructions to generate a 32 bit
LBA of 0 directly for reading the MBR. They barely fit.
YONETANI Tomokazu [Wed, 11 Feb 2009 01:06:38 +0000 (10:06 +0900)]
Mention the variables added by
3b25fbd435d9ee6d47af7128014528e2fadfb5e5
which affects the name and the location of the kernel/modules.
Sascha Wildner [Tue, 10 Feb 2009 23:04:37 +0000 (00:04 +0100)]
Sync several manpages in section 9 with reality and clean up a bit.
DEV_MODULE(9):
- adjust example
devtoname(9):
- dev_t is cdev_t now
make_dev(9):
- dev_t is cdev_t now
- make_dev() takes dev_ops as its first argument
- obtain a reference to cdev_t, so we can destroy it later
- don't forget to call dev_ops_remove() during shutdown
vcount(9):
- document count_udev(9)
Submitted-by: Stathis Kamperis <ekamperi@gmail.com>
Dragonfly-bug: <http://bugs.dragonflybsd.org/issue1269>
YONETANI Tomokazu [Tue, 10 Feb 2009 12:40:11 +0000 (21:40 +0900)]
Set the default DESTDIR to /var/vkernel when the platform is vkernel.
This saves some keystrokes as long as you use the standard installation
location for vkernel(as shown in vkernel(7)).
YONETANI Tomokazu [Tue, 10 Feb 2009 12:35:13 +0000 (21:35 +0900)]
Add two variables to adjust where to install kernel and modules.
DESTLABEL common suffix added to kernel and modules directory. for
example, `make DESTLABEL=test installkernel' installs them
as /boot/kernel.test and /boot/modules.test, respectively.
DESTKERNDIR where to install kernel and modules, /boot by default.
Add /sys/conf/kern.path.mk to define default paths for kernel and modules.
Also move KMOD* definitions to /sys/conf/kmod.mk where it really belongs.
Submitted-by: Stefan Johannesdal <stefan.johannesdal@gmail.com>
with a slight modification by me.
Sascha Wildner [Tue, 10 Feb 2009 11:35:36 +0000 (12:35 +0100)]
nrelease: Exclude distfiles as well when we tar pkgsrc.
Matthew Dillon [Tue, 10 Feb 2009 01:25:32 +0000 (17:25 -0800)]
Release Engineering: Enhance gui PFI to auto-detect xorg or curses
Add an "auto" feature for pfi_frontend and change the default to "auto".
Adjust the gui/usr/local/bin/installer script to accept "auto" and set
pfi_frontend to "xorg" or "curses" based on the existance of the DISPLAY
variable.
Sascha Wildner [Mon, 9 Feb 2009 19:30:11 +0000 (20:30 +0100)]
zoneinfo & zic(8): Sync with tzcode2009b & tzdata2009b from elsie.
* zic.c: Fix fencepost error.
* leapseconds: Updated to reflect no mid-2009 leap second.
Matthew Dillon [Mon, 9 Feb 2009 18:48:23 +0000 (10:48 -0800)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Mon, 9 Feb 2009 18:46:13 +0000 (10:46 -0800)]
HAMMER VFS: Correct a ref count leak in the recent I/O invalidation work
Correct a ref count leak in the recent I/O invalidation work. This could
lead to permanently referenced hammer_buffers and prevent reservations
from terminating.
Reported-by: Michael Neumann <mneumann@ntecs.de>
Sascha Wildner [Mon, 9 Feb 2009 18:40:02 +0000 (19:40 +0100)]
installer: Correct typo in a comment.
Matthew Dillon [Mon, 9 Feb 2009 17:31:57 +0000 (09:31 -0800)]
Release Engineering: nrelease work, more didbootstrap, include pkgsrcs, and more
* Include full sources on the DVD release (tar cz'd) - 90M
* Include the full pkgsrc tree on both the CD and DVD release (tar cz'd) - 26M
* More cleanups
* Hack fixed for .didbootstrap. clean target combinations were not resolving
the .if tests in the expected way.
Matthew Dillon [Mon, 9 Feb 2009 06:03:56 +0000 (22:03 -0800)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Mon, 9 Feb 2009 06:00:39 +0000 (22:00 -0800)]
Release Engineering: nrelease and GUI build adjustments
* Add a 'restartpkgs' target to the nrelease Makefile
* If /usr/src is not owned by root properly chown cpdup'd files to root.
* Quiet superfluous errors during certain tests
* Gui build: Do not automatically start X. Provide instructions instead.
Otherwise the DVD can take forever to boot. Remove related hacks.
* Gui build: Change the default shell to csh.
Matthew Dillon [Mon, 9 Feb 2009 02:15:04 +0000 (18:15 -0800)]
Release Engineering: nrelease build fixes - adjust MFS sizes and .zshrc
* Increase the MFS size for /var in the gui build from 16M to 128M.
The package db doesn't fit in 16M.
* Remove the '*' in the 'du' alias in the gui .zshrc
Sascha Wildner [Mon, 9 Feb 2009 01:43:29 +0000 (02:43 +0100)]
ubsa(4): Move HARDWARE section below DESCRIPTION.
Thomas Nikolajsen [Thu, 5 Feb 2009 18:07:33 +0000 (19:07 +0100)]
rc.d/nfsclient: delete obsolete flag for rpc.umntall(8)
Thomas Nikolajsen [Sun, 8 Feb 2009 22:31:12 +0000 (23:31 +0100)]
installer: change `atacontrol' to `natacontrol'
`atacontrol' doesn't work with default kernel, GENERIC,
as it uses `device nata'.
Matthew Dillon [Sun, 8 Feb 2009 20:15:52 +0000 (12:15 -0800)]
Adjust the ${ISODIR}/.didbootstrap logic a little, add needed licenses.
Adjust the bootstrap logic a bit and add the following licenses to
the release build's mk.conf so the gui packages all build: The
openmotif-license and the vim-license.
Matthias Schmidt [Sun, 8 Feb 2009 13:01:44 +0000 (14:01 +0100)]
Add new option to pkg_search
-d will download the pkg_summary(5) file for you. If it is already installed,
it get's overwritten. This is useful for users using only binary packages.
After a release -d will give you the new list of packages available for that
release.
Matthias Schmidt [Sun, 8 Feb 2009 12:04:04 +0000 (13:04 +0100)]
Shift link for 2.2
We already have packages for 2.2 on pkgbox and the mirrors.
Sascha Wildner [Sun, 8 Feb 2009 15:27:46 +0000 (16:27 +0100)]
nrelease: Restrict swap size to 8GB for now.
Thomas Nikolajsen [Sun, 8 Feb 2009 15:17:47 +0000 (16:17 +0100)]
Merge commit 'chlamydia/master'
Thomas Nikolajsen [Sun, 8 Feb 2009 14:54:26 +0000 (15:54 +0100)]
vnconfig: Fix printing of inode
Thomas Nikolajsen [Sun, 8 Feb 2009 14:41:59 +0000 (15:41 +0100)]
uuids.5: Fix error and improve wording
* UUID is 128 bit; not 64 bit
Thomas Nikolajsen [Sun, 8 Feb 2009 14:36:07 +0000 (15:36 +0100)]
jscan: Fix SYNOPSIS and sync usage()
Thomas Nikolajsen [Sun, 8 Feb 2009 14:09:18 +0000 (15:09 +0100)]
diskinfo: Add description of flags, fix SYNOPSIS and sync usage()
Thomas Nikolajsen [Sun, 8 Feb 2009 14:04:37 +0000 (15:04 +0100)]
ccdconfig: Improve markup & sync usage()
* Fix SYNOPSIS: problems in `...' use
* Sync usage() to SYNOPSIS
* Add reference to disklabel64(8)
* Add markup
* Break long command lines in examples
* Start sentence on new line
Thomas Nikolajsen [Sun, 8 Feb 2009 14:03:35 +0000 (15:03 +0100)]
strfile.8: Add missing `C' flag to SYNOPSIS.
Sascha Wildner [Sun, 8 Feb 2009 14:43:33 +0000 (15:43 +0100)]
nrelease: Add /tmp to the list of nohistory directories.
Thomas Nikolajsen [Sun, 8 Feb 2009 14:01:40 +0000 (15:01 +0100)]
sh.1: Fix markup.
A few ``...'' wasn't marked as arguments.
Matthias Schmidt [Sun, 8 Feb 2009 10:42:34 +0000 (11:42 +0100)]
Revert
4dcaa51ba4d4238e035a802067366a28527cd570
This reverts the .forward commit. I committed the code to early, there are some
bugs inside (queue handling broken and some security issues). I back this out
until we have more time to fix all the issues or rewrite some parts from scratch.
This brings dma back in a fully working state, only the .forward stuff is gone.
Tested-by: Daniel Roethlisberger <daniel@roe.ch> and me
Ok-to-back-out: corecode@
Matthias Schmidt [Sat, 7 Feb 2009 19:08:05 +0000 (20:08 +0100)]
Merge branch 'master' of git://chlamydia.fs.ei.tum.de/dragonfly
Matthew Dillon [Sun, 8 Feb 2009 10:12:25 +0000 (02:12 -0800)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Sun, 8 Feb 2009 10:08:22 +0000 (02:08 -0800)]
nrelease Makefile adjustments: Packages, newaliases, check target, etc
* Add a newaliases step which fixes the aliases.db warning when the CD boots.
* Adjust the DVD package set a bit. There are still some license issues
which need to be resolved.
* Fix the check target to detect certain required packages for building
the ISO.
* Adjust pkgsrc_bootstrap and pkgsrc_cdrecord helper targets
* Remove some stale targets which are no longer used.
Thomas Nikolajsen [Sun, 8 Feb 2009 08:04:36 +0000 (09:04 +0100)]
Fix typo.
Matthew Dillon [Sun, 8 Feb 2009 02:43:39 +0000 (18:43 -0800)]
Improve the customizeiso step to allow restarts
* Use a touch-file to determine if the bootstrap has been built, clean out
the entire pkg installation if it has not before continuing. Leave
things intact if the bootstrap has already been installed.
* Use defensive umounts and rm's as needed to allow the null mounts and
package building environment to be recreated without destroying packages
already installed.
* Allow /usr/pkgsrc to be read-only. If it is read-only then a null-mount
will be used for /usr/pkgsrc/distfiles pointing to /usr/release/distfiles.
* Add a bmake check step when building each package and only build packages
which have not already been installed.
Stathis Kamperis [Fri, 6 Feb 2009 16:33:38 +0000 (16:33 +0000)]
Polish cdev example
- dev_t is cdev_t now
- include stdlib.h needed by exit(3)
- respect the ``80 columns per line'' limit
- comply with KNF
No functional change intended
Sepherosa Ziehau [Thu, 5 Feb 2009 10:39:05 +0000 (18:39 +0800)]
pci intr config: 'j' is the pin index we want to skip.
Sepherosa Ziehau [Wed, 4 Feb 2009 10:39:27 +0000 (18:39 +0800)]
pci intr config: Fix an off by one bug
FreeBSD pci_pir.c rev1.105
We represent PCI intpin's two different ways. One is the way that the
intpin register is expressed in hardware where 0 means none, 1 means INTA,
2 INTB, etc. The other way is commonly used in loops where 0 means INTA,
1 means INTB, etc. The matchpin argument to pci_cfgintr_search() is
supposed to be the first form, but we passsed in a loop index of the
second. This fix adds one to the loop index to convert to the first form.
Reported by: Pavlin Radoslavov <pavlin@icir.org>
Aggelos Economopoulos [Thu, 5 Feb 2009 22:33:00 +0000 (00:33 +0200)]
Merge branch 'master' of git://chlamydia.fs.ei.tum.de/dragonfly
Aggelos Economopoulos [Thu, 5 Feb 2009 22:27:34 +0000 (00:27 +0200)]
kill db_print_backtrace()
Use print_backtrace() which is always available. Remove
a bunch of #ifdef DDB and #include <ddb/ddb.h> lines.
Sascha Wildner [Thu, 5 Feb 2009 22:21:18 +0000 (23:21 +0100)]
installer: Use unsigned long for holding hw.physmem's value.
The previous signed long led to the wrong calculation of default swap size
on machines with >2GB main memory.
Reported-by: corecode
Dragonfly-bug: <http://bugs.dragonflybsd.org/issue1257>
Sascha Wildner [Thu, 5 Feb 2009 11:38:45 +0000 (12:38 +0100)]
nrelease: Whitespace.
Sascha Wildner [Thu, 5 Feb 2009 11:37:47 +0000 (12:37 +0100)]
nrelease: Re-add cvsup package URL which I accidentally removed.
Simon Schubert [Thu, 5 Feb 2009 09:55:24 +0000 (10:55 +0100)]
tcsh: export PATH if using the default setting
Simon Schubert [Thu, 5 Feb 2009 09:37:45 +0000 (10:37 +0100)]
define _PATH_DEFPATH to be reasonable, including sbins and /usr/pkg
sh already had this path set as default, but only sh. Now everybody
using _PATH_DEFPATH will get the right thing.
Simon Schubert [Thu, 5 Feb 2009 09:34:50 +0000 (10:34 +0100)]
sh: export the default PATH instead of keeping it private
sh specifies a default setting for PATH if none is set in the
environment, but up to now it didn't export this setting. This leads to
strange effects, like sh being able to execute binaries in /usr/sbin,
but which(1) not being able to find them. Export PATH to create more
symmetry and harmony.
Simon Schubert [Thu, 5 Feb 2009 09:33:55 +0000 (10:33 +0100)]
sh: use _PATH_DEFPATH instead of hand-rolling the default path
Matthew Dillon [Thu, 5 Feb 2009 06:19:42 +0000 (22:19 -0800)]
MFC FreeBSD/1.49 - fix misuse of system() and failure if /usr is not mounted.
Suggested-by: Jeremy Chadwick <jdc@parodius.com>
Matthew Dillon [Thu, 5 Feb 2009 05:20:49 +0000 (21:20 -0800)]
Release Engineering: Add ds macros for 2.2 and 2.3.
Matthew Dillon [Thu, 5 Feb 2009 05:04:04 +0000 (21:04 -0800)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Thu, 5 Feb 2009 05:03:11 +0000 (21:03 -0800)]
Release Engineering: Update master to 2.3.0-DEVELOPMENT
Simon Schubert [Wed, 4 Feb 2009 22:17:54 +0000 (23:17 +0100)]
nrelease: prevent leaking of env vars into pkgsrc build
At least TARGET_ARCH, which is used by the release build, is causing
havoc for packages built by gmake, as gmake appends ${TARGET_ARCH} to
the ${COMPILE.c} compiler command line.
To be on the safe side, clear the environment completely when calling
into the pkgsrc build.
Sascha Wildner [Wed, 4 Feb 2009 22:10:47 +0000 (23:10 +0100)]
nrelease: Limit the volid to 32 chars, when building from a dirty state.
Sascha Wildner [Tue, 3 Feb 2009 22:06:00 +0000 (23:06 +0100)]
mount(8): Start sentence on new line.
Matthew Dillon [Tue, 3 Feb 2009 21:58:12 +0000 (13:58 -0800)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Tue, 3 Feb 2009 21:53:54 +0000 (13:53 -0800)]
Suppress warnings if /etc/fstab is missing when generating output for
mount -p. Generate reasonable defaults to fill in missing pieces.
Submitted-by: Stathis Kamperis <ekamperi@gmail.com> added the comment:
Simon Schubert [Fri, 30 Jan 2009 15:15:44 +0000 (15:15 +0000)]
GENERIC: replace tab with space to line up even when commenting out
Sascha Wildner [Tue, 3 Feb 2009 17:24:48 +0000 (18:24 +0100)]
Add a vcount(9) manual page.
Submitted-by: Stathis Kamperis <ekamperi@gmail.com>
Dragonfly-bug: <http://bugs.dragonflybsd.org/issue1266>
Sascha Wildner [Tue, 3 Feb 2009 08:26:17 +0000 (09:26 +0100)]
Silence 3 warnings in libc.
Matthew Dillon [Sun, 1 Feb 2009 08:38:23 +0000 (00:38 -0800)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Sun, 1 Feb 2009 08:28:30 +0000 (00:28 -0800)]
Comment out the session line for PAM's xdm config. This works around
a widely known bug in xdm where a PAM callback can cause xdm to access
a bogus widget pointer after the widget has been destroyed.
With-help-from: Sephe
Hasso Tepper [Sun, 1 Feb 2009 07:33:12 +0000 (09:33 +0200)]
Make XPT_SCSI_IO path quiet again.
Now when problems with SATA ATAPI devices are resolved, it's not
helpful any more.
Matthew Dillon [Sun, 1 Feb 2009 06:53:38 +0000 (22:53 -0800)]
HAMMER VFS - Fix I/O invalidation collision assertion
Fix more issues with the hammer_reserve subsystem.
* Fix a broken base_offset calculation which had reversed the zone
and offset fields in a HAMMER_ZONE_ENCODE() call. The broken
base_offset caused hammer_del_buffers() to scan the wrong offset
range, leaving conflicting buffer cache buffers intact.
The resulting conflict caused a KKASSERT(LIST_FIRST(&bp->b_dep) == NULL)
to fail later on.
* Hold a reference through a potentially blocking operation in
hammer_blockmap_alloc().
* Record the proper zone in a hammer_reserve() structure created via
hammer_blockmap_free(). Without this the I/O invalidation code
cannot locate conflicting hammer_buffer structures.
* hammer_reserve_setdelay_offset() was not actually placing the
hammer_reserve() structure in the delay queue, allowing it to
be disposed of too early.
Sascha Wildner [Sat, 31 Jan 2009 23:35:38 +0000 (00:35 +0100)]
make(1) tests: Remove remaining CVS IDs from the tests to make them run again.
Sascha Wildner [Sat, 31 Jan 2009 22:57:25 +0000 (23:57 +0100)]
Remove some unexpanded $DragonFly$ IDs in our tree.
Sascha Wildner [Sat, 31 Jan 2009 22:51:37 +0000 (23:51 +0100)]
make(1) tests: Messing with the CVS ID is no longer needed.
Sascha Wildner [Sat, 31 Jan 2009 22:13:20 +0000 (23:13 +0100)]
rpc_svc_calls(3): Remove __svc_getcallercreds() documentation.
FreeBSD removed the macro but forgot to adjust the manpage.
Sascha Wildner [Sat, 31 Jan 2009 21:06:42 +0000 (22:06 +0100)]
make.conf: Bring back WARNS_WERROR to build with -Werror if WARNS is set.
In-discussion-with: pavalos
Sascha Wildner [Sat, 31 Jan 2009 18:41:02 +0000 (19:41 +0100)]
Typo fixing round.
Sascha Wildner [Sat, 31 Jan 2009 13:05:39 +0000 (14:05 +0100)]
installer: Fix reading disk descriptions when booting verbose.
This fixes issue1256.
Reported-by: corecode
Sascha Wildner [Sat, 31 Jan 2009 07:21:16 +0000 (08:21 +0100)]
installer: Remove .gitignore files.
Sascha Wildner [Sat, 31 Jan 2009 09:17:02 +0000 (10:17 +0100)]
manpages: Split DIAGNOSTICS into EXIT STATUS and DIAGNOSTICS.
DIAGNOSTICS shall have information about diagnostic messages while
exit code information (.Ex macro) goes to EXIT STATUS.
Inspired-by: FreeBSD
Sascha Wildner [Sat, 31 Jan 2009 08:17:04 +0000 (09:17 +0100)]
priv(9): We don't have prison_priv_check() yet so comment out some things.
Sepherosa Ziehau [Thu, 22 Jan 2009 09:13:20 +0000 (17:13 +0800)]
syncache_socket(): fix abort path by calling pru_abort directly
On syncache_socket() abort path, the inp's {fport,lport,faddr,laddr}
is either incomplete or half destroyed, which causes tcp_soport()
mapping the socket to a wrong CPU. However, in syncache code, inp
is only accessed or created by its owner CPU, i.e. we are already on
the inp's owner CPU during the socket aborting.
Add soabort_oncpu(), which calls pru_abort directly. Use it on
syncache_socket() abort path.
Matthew Dillon [Fri, 30 Jan 2009 17:05:17 +0000 (09:05 -0800)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Fri, 30 Jan 2009 16:46:38 +0000 (08:46 -0800)]
Fix a string parsing bug when trying to run hammer pfs-master on a
subdirectory of the root directory, e.g. "hammer pfs-master /fubar".
Reported-by: "Simon 'corecode' Schubert" <corecode@fs.ei.tum.de>
Sascha Wildner [Fri, 30 Jan 2009 09:51:12 +0000 (10:51 +0100)]
Put the repo version in the ISO's volid.
Sascha Wildner [Fri, 30 Jan 2009 08:44:30 +0000 (09:44 +0100)]
nrelease: Use a separate workdir to keep PKGSRC_PATH clean.
Peter Avalos [Fri, 30 Jan 2009 03:06:11 +0000 (19:06 -0800)]
Fix capital letter in the middle of a sentence.
Submitted-by: Stathis Kamperis <ekamperi@gmail.com>
Sascha Wildner [Thu, 29 Jan 2009 21:58:11 +0000 (22:58 +0100)]
jail(8): Fix path in the example.
Sascha Wildner [Thu, 29 Jan 2009 18:57:06 +0000 (19:57 +0100)]
usbhid(3): Re-commit some fixes that were undone in the FreeBSD sync.
Simon 'corecode' Schubert [Thu, 29 Jan 2009 18:21:03 +0000 (18:21 +0000)]
cvs: obey read only fs setting also for the val-tags file
Simon Schubert [Thu, 29 Jan 2009 09:34:03 +0000 (10:34 +0100)]
nata: process interrupt even if busmaster active bit is set
The nata driver previously assumed that an interrupt which has the
interrupt and active bits set in the busmaster status register is not a
valid/completed interrupt. In these situations if would return without
acknowledging the irq, leading to an interrupt livelock situation.
This condition of INTERRUPT|ACTIVE occured on two different machines
both equipped with different versions of an ICH SATA controller, when
accessing an attached SATA ATAPI (DVD) drive by TSScorp (Samsung
OEM/bulk). Specifically, natapicam seems to trigger this condition on
attach. It seems that the drive is unhappy with the INQUIRY command
sent by natapicam. Related artifacts appear when trying to access a
video DVD (possibly due to missing region/css key).
Under the hood it seems that the drive reports an error condition
(visible in the ALT/STATUS register), but does not clear the ACTIVE bit
of the busmaster status. As a result the nata driver could be waiting
for the completion of the transfer, while the SATA busmaster would keep
the interrupt line asserted.
We work around this issue by following the precedent of other operating
systems and ignoring the ACTIVE bit if the INTERRUPT bit is set.
Specifically, we wait a little bit to accomodate for possibly faulty
hardware which might assert the irq before the transfer is completely
finished. Nevertheless, we do not reject interrupt proecessing if the
INTERRUPT bit is set in the busmaster status register.
Sascha Wildner [Thu, 29 Jan 2009 06:22:24 +0000 (07:22 +0100)]
nrelease: Switch to building pkgsrc packages from source.
The pkgsrc tree that shall be used must be specified via PKGSRC_PATH.
So far the LiveCD build has been successfully tested. The LiveDVD might
still need some love, but I'm committing this now so work on the release
can begin.
Still on the TODO is using binary packages optionally.
Requested-by: dillon
Sascha Wildner [Wed, 28 Jan 2009 19:47:34 +0000 (20:47 +0100)]
Fix format order.
Reported-by: corecode
Matthew Dillon [Tue, 27 Jan 2009 01:36:19 +0000 (17:36 -0800)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Tue, 27 Jan 2009 01:35:34 +0000 (17:35 -0800)]
HAMMER VFS - Add missing bqrelse() in last patch.
This fixes a deadlock introduced by the last patch due to a missing
bqrelse() call.
Simon Schubert [Fri, 23 Jan 2009 00:23:27 +0000 (01:23 +0100)]
Revert "Update the atime when mmap'ing a file"
This reverts commit
8e7ceff7f24f4bd5eae767c8c091e696d490dd58.
Better code committed in
a6e41612e54fa7b0a616c8f8d14446bab206d030
Simon Schubert [Fri, 23 Jan 2009 00:22:25 +0000 (01:22 +0100)]
hda: disable verbose output - too chatty even for bootverbose
Simon Schubert [Thu, 22 Jan 2009 20:40:53 +0000 (21:40 +0100)]
nata: revert
6e40a325 = don't use tsleep in ata_udelay
ata_udelay might be called from softclock context due to a request
timeout. However we must not tsleep in softclock. This restores the
pre-
6e40a325 behavior.
Matthew Dillon [Mon, 26 Jan 2009 06:34:42 +0000 (22:34 -0800)]
HAMMER VFS - Reorganize the use of hammer_del_buffers(), fix rare deadlocks
The previous attempt at fixing a deadlock due to a hammer_buffer being
held cached across long-blocking procedures failed to catch all cases.
Implement a more complete and less invasive fix. These buffers are
typically held in the hammer_cursor structure.
Instead of trying to release such cached data buffers proactively,
we instead allow hammer_del_buffers() to return a temporary failure when
called from the reservation code. If hammer_del_buffers() is unable to
completely eradicate underlying buffers due to them being referenced via
the above cache entities the related hammer_reserve structure is simply
re-added to the delayed-release queue.
The related big-block are thus not reused until all such cached entites
have been dereferenced. The relaxed requirements on the cached entities
removes the possibility of a deadlock in the invalidation code and also
simplifies hammer_io_inval()'s handling of the hammer_buffer<->buffer_cache
coupling case.