Matthew Dillon [Wed, 12 Aug 2009 04:53:45 +0000 (21:53 -0700)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Wed, 12 Aug 2009 04:52:58 +0000 (21:52 -0700)]
AMD64 - Sync additional modifications to amd64 test Makefile
* ad1 becomes ad0 in the qemu-run amd64 kernel.
Matthew Dillon [Wed, 12 Aug 2009 04:48:04 +0000 (21:48 -0700)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly into amd64
Matthew Dillon [Wed, 12 Aug 2009 04:45:26 +0000 (21:45 -0700)]
AMD64 - Sync additional modifications so amd64 kernel & world builds on i386.
Matthew Dillon [Wed, 12 Aug 2009 04:43:23 +0000 (21:43 -0700)]
AMD64 - Sync machine-dependent bits from smtms.
Submitted-by: Jordan Gordeev <jgordeev@dir.bg>
Matthew Dillon [Wed, 12 Aug 2009 04:41:31 +0000 (21:41 -0700)]
AMD64 - Sync machine-independant bits from smtms.
Submitted-by: Jordan Gordeev <jgordeev@dir.bg>
Matthew Dillon [Wed, 12 Aug 2009 00:36:53 +0000 (17:36 -0700)]
CISS - Fake up SYNCHRONIZE_CACHE by default.
Reported-by: =?ISO-8859-1?Q?Gy=F6rgy_Vilmos?= <vilmos.gyorgy@gmail.com>
Matthew Dillon [Wed, 12 Aug 2009 00:09:59 +0000 (17:09 -0700)]
mount - Finish refactoring mount -v and MOUNTCTL_MOUNTFLAGS
* mount -v now uses MOUNTCTL_MOUNTFLAGS to retrieve the mount flags
* The kernel MOUNTCTL_MOUNTFLAGS implementation now makes it easy for
a VFS (like HAMMER) to append its own flags.
Submitted-by: Antonio Huete
Matthew Dillon [Tue, 11 Aug 2009 17:07:49 +0000 (10:07 -0700)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Tue, 11 Aug 2009 17:06:01 +0000 (10:06 -0700)]
disklabel* - Make disk and filesystem types case insensitive.
* e.g. allow 'hammer' or 'HAMMER' to be specified when configuring a
filesystem type for a partition. When displayed the proper case will be
used.
This reduces confusion for people trying to partition a disk manually.
Alex Hornung [Tue, 11 Aug 2009 16:04:33 +0000 (17:04 +0100)]
DEVFS - fix probing of label for MBR compat slice; cleanup whitespaces
* Also probe for labels if no real mbr but only a compat slice was
"found" by subr_diskmbr. This avoided probing of the compat slice
as for example described in vkernel(7)
* Clean up whitespaces
Reported-By: YONETANI Tomokazu
Alex Hornung [Tue, 11 Aug 2009 14:05:16 +0000 (15:05 +0100)]
devfsctl - Completely rewrite the parser;separate rule representation
* Take over Simon "corecode" Schubert's changes to separate the rule
representation in userland from the kernel representation, adapting
it to the new parser.
* Completely rewrite the parser to be more maintainable, support a
clearer and better rule format and give better warnings/errors.
Sascha Wildner [Tue, 11 Aug 2009 14:45:35 +0000 (16:45 +0200)]
compilers.conf(5): Kick gcc44 from the config, now that it's in base.
Sascha Wildner [Tue, 11 Aug 2009 14:28:35 +0000 (16:28 +0200)]
boot2: prevent use of DEBUG_FLAGS
When compiling with -g, gcc44 embeds debugging data into the boot2
binary which can not be stripped by objcopy. This results in boot2
being too large to fit into the boot sector.
Generally prevent changes to the fragile boot2 build by ignoring
DEBUG_FLAGS completely.
Sascha Wildner [Tue, 11 Aug 2009 13:38:11 +0000 (15:38 +0200)]
Add some documentation for fdevname() and fdevname_r() (thanks, alexh).
Sascha Wildner [Tue, 11 Aug 2009 13:37:16 +0000 (15:37 +0200)]
stdlib.h: Adjust indentation a bit.
Alex Hornung [Tue, 11 Aug 2009 11:07:16 +0000 (12:07 +0100)]
ttyname(3) - Return EBADF if the fd passed is invalid
* Return EBADF instead of ENOTTY if the passed file descriptor is not
valid to conform with standard.
Alex Hornung [Tue, 11 Aug 2009 09:29:45 +0000 (10:29 +0100)]
fdevname - Set errno if fdevname_r call fails; don't prepend /dev
* Return EINVAL if the fd does not belong to a character device and
return EBADF if the fd is not valid.
* If fdevname_r returns non-null when called from fdevname, set errno
in fdevname to whatever *_r returned.
* Don't prepend "/dev/" string to the returned buffer, just put in
the device name itself, without absolute path.
Reported-By: Sascha Wildner
Alex Hornung [Tue, 11 Aug 2009 09:19:57 +0000 (10:19 +0100)]
ttyname(3) - Set errno if ttyname_r call fails
* If ttyname_r returns non-null when called from ttyname, set errno
in ttyname to whatever *_r returned.
This also makes our ttyname comply with the standard.
Reported-By: Sascha Wildner
Sascha Wildner [Tue, 11 Aug 2009 09:01:31 +0000 (11:01 +0200)]
installer: GC some dead code.
Sascha Wildner [Tue, 11 Aug 2009 09:00:36 +0000 (11:00 +0200)]
installer: use cmd_name()
Alex Hornung [Mon, 10 Aug 2009 19:18:03 +0000 (20:18 +0100)]
DEVFS - Modification of clone behaviour; general cleanup; xtime
* Use default_dev_ops by default for make_autoclone_dev so that opens
and other operations on it fail when it fails as a clone handler.
* Accept NULL as argument for bimtap for make_autoclone_dev, so that
there's no explicit need for a clone bitmap to use this function.
* Add destroy_autoclone_dev as a counterpart to make_autoclone_dev.
* Const-ify a bunch of char *
* Use kstrdup instead of kmalloc + memcpy magic at a few places where
this makes sense.
* Rework atime/mtime/ctime as advised by Simon Schubert on IRC.
* Clean up devfs_clone by getting rid of old code and return NULL
if cloning fails, else return the cloned device.
* Sync up all pending devfs operations just before calling into the
clone handler, so that issues with device deletion/creation are
dealt with appropriately.
* Change cloning in devfs_spec_open to match devfs_clone changes and
also check the return value of devfs_create_device_node, as it can
return NULL. If cloning fails, the cloning device will be opened,
which will have the default_ops and hence will fail returning
ENODEV.
* Cosmetic changes
Suggested-By: Simon "corecode" Schubert
Sascha Wildner [Tue, 11 Aug 2009 08:12:40 +0000 (10:12 +0200)]
'options DEVFS' isn't anymore, so remove useless VKERNEL_DEVFS config.
Sascha Wildner [Tue, 11 Aug 2009 02:19:14 +0000 (04:19 +0200)]
List NO_GCC44 in make.conf and add some lines to the manpage for it.
Sascha Wildner [Tue, 11 Aug 2009 01:59:10 +0000 (03:59 +0200)]
vkd(4): Comment out an unused variable (just like the code).
Sascha Wildner [Tue, 11 Aug 2009 01:58:20 +0000 (03:58 +0200)]
i4b(4): Remove a declaration which is no longer needed.
Matthew Dillon [Tue, 11 Aug 2009 02:13:58 +0000 (19:13 -0700)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Tue, 11 Aug 2009 02:11:35 +0000 (19:11 -0700)]
nrelease - adjust usb disk-key image generator
* Add an explicit vfs.root.mountfrom line in loader.conf to point
the boot loader at da8.
* Makefile cleanup. Remove the one-huge-sh line and properly dismount and
unconfigure any left over VN from previous attempts.
Nicolas Thery [Mon, 10 Aug 2009 21:29:52 +0000 (23:29 +0200)]
crget(): use M_ZERO instead of bzero()
Thomas E. Spanjaard [Mon, 10 Aug 2009 20:20:51 +0000 (20:20 +0000)]
gcc44: fix buildworld on amd64
Matthew Dillon [Mon, 10 Aug 2009 16:32:13 +0000 (09:32 -0700)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Mon, 10 Aug 2009 16:30:27 +0000 (09:30 -0700)]
USB - Clean up address setting and softc allocations
* Zero out the softc structure being allocated for ehci and ohci.
* When establishing a new connection and allocating an address for
further communications, disconnect and reconnect the default pipe
instead of just ripping up the underlying address.
Research-done-by: Rumko <rumcic@gmail.com>
Partially-taken-from: FreeBSD
Matthew Dillon [Mon, 10 Aug 2009 16:27:42 +0000 (09:27 -0700)]
EHCI - Fix early probing problems.
* When polling is used an actual interrupt will clear pending status to
prevent an interrupt livelock. However, if this occurs the polling code
can wind up believing that no status change has occured.
Add the cleared pending status to sc->sc_dintrs and then reincorporate
into the status check when polling or running a normal interrupt.
* Code cleanup: NULL out some fields on termination.
Sascha Wildner [Mon, 10 Aug 2009 14:17:48 +0000 (16:17 +0200)]
shlock(1): Remove a wrong free() of a stack variable.
Sascha Wildner [Mon, 10 Aug 2009 14:15:02 +0000 (16:15 +0200)]
ahci.4: Oops forgot one .Nm
Sascha Wildner [Mon, 10 Aug 2009 14:08:26 +0000 (16:08 +0200)]
development.7: Improve markup a bit, bump .Dd.
Sascha Wildner [Mon, 10 Aug 2009 14:07:38 +0000 (16:07 +0200)]
ahci.4: Improve markup, begin sentences on a new line, bump .Dd.
Sascha Wildner [Mon, 10 Aug 2009 14:05:14 +0000 (16:05 +0200)]
chmod.2/mountctl.2: Remove trailing whitespace.
Sascha Wildner [Mon, 10 Aug 2009 08:30:20 +0000 (10:30 +0200)]
hexdump: always stat input when skipping
The previous code would not fstat(2) the input, if operating on stdin.
Still, it would use the uninitialized struct stat to decide whether to
seek in the input or not. This could either lead to hexdump reading the
skipped input, instead of directly seeking, or, worse, trying a seek
even though this not being possible.
Fix this issue by always fstat()ing the input.
Submitted-by: corecode
Sascha Wildner [Mon, 10 Aug 2009 08:23:04 +0000 (10:23 +0200)]
ee(1): Fix some wrong usage of NULL.
Matthew Dillon [Mon, 10 Aug 2009 06:17:32 +0000 (23:17 -0700)]
NFS - Cleanly unmount NFS on halt/shutdown/reboot
* Abort any in-transit RPCs, disallow new RPCs, and disconnect the
socket (aka if TCP) when doing a forced NFS unmount.
This fixes left over TCP sockets when doing a normal halt/shutdown/reboot.
* Related user processes will also unblock, but the unmount code will
still block trying to look up the path and this commit does not quite
solve that problem.
Matthew Dillon [Mon, 10 Aug 2009 03:25:57 +0000 (20:25 -0700)]
DEVFS - Make synchronization prior to mountroot more robust
* A single disk_config() + devfs_config() isn't enough, call them twice.
(tested with TWA driver which seems to trigger the problem).
Not exactly sure why one pass doesn't do the job but I think
pending devfs commands might trigger additional disk probe
operations.
Matthew Dillon [Mon, 10 Aug 2009 00:21:18 +0000 (17:21 -0700)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Mon, 10 Aug 2009 00:20:34 +0000 (17:20 -0700)]
DEVTAB - Add support in umount, fsck, and hammer, and add sample /etc/devtab.
Nicolas Thery [Sun, 9 Aug 2009 21:01:42 +0000 (23:01 +0200)]
mknod(2) man: fix incorrect include
Matthew Dillon [Sun, 9 Aug 2009 18:22:30 +0000 (11:22 -0700)]
AMD64 - Fix prototypes for atomic_cmpset_long()
Reported-by: smtms
Matthew Dillon [Sun, 9 Aug 2009 18:15:35 +0000 (11:15 -0700)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Sun, 9 Aug 2009 18:15:14 +0000 (11:15 -0700)]
AMD64 - Fix atomic_cmpset_long()
Matthew Dillon [Sun, 9 Aug 2009 18:14:33 +0000 (11:14 -0700)]
AMD64 - Fix 64 bit buildworld, missing compiler options in sys/boot
Matthew Dillon [Sun, 9 Aug 2009 17:23:57 +0000 (10:23 -0700)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Nicolas Thery [Sun, 9 Aug 2009 17:30:51 +0000 (19:30 +0200)]
improve a bit chmod(2) man page
Nicolas Thery [Sun, 9 Aug 2009 17:33:23 +0000 (19:33 +0200)]
add fchownat(2) system call
Add also libc_r wrapper and man page.
Matthew Dillon [Sun, 9 Aug 2009 17:21:22 +0000 (10:21 -0700)]
GCC - Add NO_GCC44 feature and use it in test/amd64/Makefile
* Feature added primarily so we can get amd64 buildworlds working again,
until gcc-4.4 starts building with it.
Thomas E. Spanjaard [Sun, 9 Aug 2009 11:13:48 +0000 (11:13 +0000)]
gcc44: adjust README.DELETED
Thomas E. Spanjaard [Sun, 9 Aug 2009 11:09:44 +0000 (11:09 +0000)]
gcc44/gmp: add amd64 Makefile for gmp
Thomas E. Spanjaard [Sun, 9 Aug 2009 10:59:53 +0000 (10:59 +0000)]
Merge branch 'vendor/GCC44'
Thomas E. Spanjaard [Sun, 9 Aug 2009 10:58:57 +0000 (10:58 +0000)]
gcc44: import forgotten file
Simon Schubert [Sat, 8 Aug 2009 23:30:50 +0000 (01:30 +0200)]
nrelease: do not mount devfs for build
We should not rely on devfs being present on the build host, since it
might be running an older kernel (e.g. release). Instead of mounting
/dev into the chroot via devfs, use a null mount of the host /dev.
Suggested-by: dillon@
Simon Schubert [Sat, 8 Aug 2009 22:49:20 +0000 (00:49 +0200)]
gcc44: adjust README.DELETED
Simon Schubert [Sun, 9 Aug 2009 08:32:29 +0000 (10:32 +0200)]
Merge branch 'vendor/GCC44'
Merge commit '
94905568eb324c454ac84bbe36580efad3a0a0a7' into HEAD
Simon Schubert [Tue, 4 Aug 2009 14:27:00 +0000 (16:27 +0200)]
hammer util: rename function to match behavior
cleanup_snapshots actually creates a snapshot
Simon Schubert [Tue, 4 Aug 2009 14:23:57 +0000 (16:23 +0200)]
hammer util: avoid running concurrent cleanups
If hammer cleanup gets called more frequently than the cleanup runs
take, they build up, all trying to run in parallel. Prevent this by
locking the config file while cleanup is running. If cleanup finds the
config file locked for a PFS, it will skip processing.
Simon Schubert [Thu, 23 Jul 2009 14:09:15 +0000 (16:09 +0200)]
pcm: perform stereo volume mixing for vchans
The vchan code before only performed an average volume mix for both
channels, thus not allowing to pan between left and right vchan channel.
Now we perform actual stereo volume mixing for left and right channel.
Matthew Dillon [Sun, 9 Aug 2009 02:35:19 +0000 (19:35 -0700)]
bioq - Add a sysctl kern.bioq_barrier to control bioqdisksort()'s sorting.
Control how well sorted we want bioqdisksort() to be. The value specified
in kern.bioq_barrier is the number of bio's that will be sorted until a
barrier is inserted, starting a new set. The default value is 16.
This allows testing filesystem throughput under varying conditions.
Suggested-by: Jeffrey Hsu <hsu@dragonflybsd.org>
Matthew Dillon [Sun, 9 Aug 2009 02:34:07 +0000 (19:34 -0700)]
tsleep - close a tsleep/wakeup race.
* Interlock the tsleep prior to releasing the current thread to close a
sleep/wakeup race.
Matthew Dillon [Sun, 9 Aug 2009 02:30:39 +0000 (19:30 -0700)]
PIPE - Fix bugs in pipe_shutdown(), and multiple readers.
* Pipe_shutdown() was setting the wrong EOF bits.
* When multiple readers are present pipe_end_uio() was trying test
whether an unsigned value was negative, resulting in an endless
tsleep(). Make the fields in question to signed.
Reported-by: Numerous people.
Simon Schubert [Sat, 8 Aug 2009 22:47:05 +0000 (00:47 +0200)]
gcc44: import forgotten file
Matthew Dillon [Sat, 8 Aug 2009 22:42:16 +0000 (15:42 -0700)]
DEVFS - Cleanup. Remove svn_read() / svn_write() and specvnode_fileops.
Matthew Dillon [Sat, 8 Aug 2009 22:40:57 +0000 (15:40 -0700)]
fifos - fix broken mkfifo style fifos.
* open()s of mkfifo style fifos with O_TRUNC were being disallow.
Allow the flag (it is harmless and does nothing).
Matthew Dillon [Sat, 8 Aug 2009 22:09:51 +0000 (15:09 -0700)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Sat, 8 Aug 2009 22:07:40 +0000 (15:07 -0700)]
bioqdisksort - fixes to avoid starvation
Long chains of pipelined write I/O were being sorted in front of other
requests. Due to the pipelining these other requests would wind up
getting starved virtually permanently.
Prevent starvation by forcing one out of every 16 BIOs to be ordered.
This fixes issues with HAMMER which tends to have more of an absolute
ordering of meta data verses data then UFS.
Stathis Kamperis [Fri, 24 Jul 2009 10:13:23 +0000 (13:13 +0300)]
printenv(1): Rework how env vars are looked up.
Before:
~% env foo=bar=lala printenv foo
bar=lala
~% env foo=bar=lala printenv foo=bar
lala
~%
After:
~% env foo=bar=lala printenv foo
bar=lala
~% env foo=bar=lala printenv foo=bar
~%
Dragonfly-bug: <http://bugs.dragonflybsd.org/issue1341>
Matthew Dillon [Sat, 8 Aug 2009 17:31:18 +0000 (10:31 -0700)]
DEVFS - More GPT fixes. Fix serial number aliases for slice 0.
Matthew Dillon [Sat, 8 Aug 2009 17:10:48 +0000 (10:10 -0700)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Alex Hornung [Sat, 8 Aug 2009 16:42:02 +0000 (17:42 +0100)]
DEVFS - Factor out recursive code into iterate_topology
* Factor out the recursive code which scans through all nodes in
the topology into a function, devfs_iterate_topology, which takes
a callback as argument, which is called for every node.
* Create new callback functions for all functions that were recursive
before, containing the code that was executed for each node.
* Change all callers of recursive functions to call
devfs_iterate_topology with the respective callback function
instead of having their own recursive functions.
Matthew Dillon [Sat, 8 Aug 2009 17:10:34 +0000 (10:10 -0700)]
DEVFS - probe for disklabels in GPT slices.
Matthew Dillon [Sat, 8 Aug 2009 17:10:14 +0000 (10:10 -0700)]
DEVFS - Cleanup pass.
Matthew Dillon [Sat, 8 Aug 2009 16:45:45 +0000 (09:45 -0700)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Sat, 8 Aug 2009 16:45:15 +0000 (09:45 -0700)]
DEVFS - Refactor the loop in disk_probe() to fix GPT partition #0 (s0).
Alex Hornung [Sat, 8 Aug 2009 14:38:40 +0000 (15:38 +0100)]
DEVFS - (temporarily) fix LINT kernel build
Fix building LINT by removing a stray dev_ops_add in fb and
temporarily removing vinum from the build.
Sepherosa Ziehau [Sat, 8 Aug 2009 05:22:31 +0000 (13:22 +0800)]
ACPICA-UNIX: Building fixup
Submitted-by: Alexander Polakov<polachok@gmail.com> w/ modification
Sepherosa Ziehau [Sat, 8 Aug 2009 13:13:47 +0000 (21:13 +0800)]
Merge branch 'vendor/ACPICA-UNIX'
Stathis Kamperis [Sat, 8 Aug 2009 05:14:52 +0000 (08:14 +0300)]
cproj.3: Add missing man page.
Sepherosa Ziehau [Sat, 8 Aug 2009 05:00:47 +0000 (13:00 +0800)]
Import acpica-unix-
20090521
Matthew Dillon [Sat, 8 Aug 2009 00:37:19 +0000 (17:37 -0700)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Sat, 8 Aug 2009 00:35:36 +0000 (17:35 -0700)]
DEVFS - Add required vget when disassociating a vnode from a node.
* We have to protect the vnode when disconnecting the rdev and
disassociating it from the devfs_node.
* This also has the side effect of properly cycling the vnode refs
so the namecache gets cleaned up when a node is removed. Otherwise
the namecache entry will be left intact, resolved to the now dead
vnode.
Stathis Kamperis [Fri, 7 Aug 2009 17:25:27 +0000 (20:25 +0300)]
cacosh(3): Return correct principal value.
Discussed with Matthias Drochner:
http://mail-index.netbsd.org/source-changes/2009/08/03/msg223605.html
Alex Hornung [Fri, 7 Aug 2009 21:32:29 +0000 (22:32 +0100)]
DEVFS - delete rule-created aliases on rule reset
* add a new devfs flag, DEVFS_RULE_CREATED, which is set for alias
nodes created from within the rule system.
* On rule reset, if a node is an alias and has the flag set, it is
unlinked and freed.
Sascha Wildner [Sun, 2 Aug 2009 20:58:52 +0000 (22:58 +0200)]
directory.3: Fix wrong date.
Sascha Wildner [Sat, 1 Aug 2009 16:37:08 +0000 (18:37 +0200)]
stat.2: Document fstatat().
While here, clean up a bit (more is needed, though).
Most-of-it-taken-from: FreeBSD
Sascha Wildner [Sun, 2 Aug 2009 20:39:11 +0000 (22:39 +0200)]
directory.3: Add some lines about the new fdopendir() function.
Sascha Wildner [Sun, 2 Aug 2009 19:38:46 +0000 (21:38 +0200)]
ifconfig.8: Remove empty line.
Sascha Wildner [Sun, 2 Aug 2009 16:39:56 +0000 (18:39 +0200)]
mdoc cleanup: Add missing .Os
Simon Schubert [Thu, 6 Aug 2009 14:48:34 +0000 (16:48 +0200)]
development.7: add descrption on how to perform vendor imports
Alex Hornung [Fri, 7 Aug 2009 20:52:24 +0000 (21:52 +0100)]
devfsctl - Cleanup and several bugfixes
* split lines longer than 80 columns
* Moved TAILQ_INSERT* into get_rule to avoid a problem which
caused groups to also be inserted and kept in the rule
list
* Fix several bugs related to config file parsing, including:
- a bug in the argument checker, which caused a SEGFAULT
- a bug that would cause lines beginning with whitespaces
to duplicate their first real token
- an incorrect check for an empty line, which caused a
warning about an incorrect verb instead of ignoring
the line
Matthew Dillon [Fri, 7 Aug 2009 20:13:33 +0000 (13:13 -0700)]
vkernel build suite - add README describing network bridging.
Matthew Dillon [Fri, 7 Aug 2009 20:09:04 +0000 (13:09 -0700)]
Merge branch 'master' of ssh://crater.dragonflybsd.org/repository/git/dragonfly
Matthew Dillon [Fri, 7 Aug 2009 20:08:39 +0000 (13:08 -0700)]
vkernel build suite - change -I network from /dev/vknet to /var/run/vknet
Alex Hornung [Fri, 7 Aug 2009 20:23:41 +0000 (21:23 +0100)]
subr_disk - don't show 0-size slices
When probing the disk, check for 0-size slices and avoid calling
make_dev for them.
Reported-By: Simon "corecode" Schubert
Simon Schubert [Fri, 7 Aug 2009 19:58:53 +0000 (21:58 +0200)]
kern_memio: don't treat any memio device as /dev/zero
We actually have to compare zerodev to the passed argument in
iszerodev(), and not compare it to NULL.