platform/upstream/dracut.git
12 years ago95iscsi: add iscsiuio
Harald Hoyer [Wed, 9 Nov 2011 12:58:54 +0000 (13:58 +0100)]
95iscsi: add iscsiuio

install iscsiuio
start iscsiuio
stop iscsiuio before switching root

12 years ago95iscsi/parse-iscsiroot.sh: force load a lot of kernel modules
Harald Hoyer [Wed, 9 Nov 2011 12:52:55 +0000 (13:52 +0100)]
95iscsi/parse-iscsiroot.sh: force load a lot of kernel modules

A lot of iSCSI driver are not autoloaded, so we have to manually
force load them.

12 years ago95iscsi/iscsiroot: use initiator-name from ibft firmware if unset
Harald Hoyer [Wed, 9 Nov 2011 12:48:08 +0000 (13:48 +0100)]
95iscsi/iscsiroot: use initiator-name from ibft firmware if unset

If initiator-name is unset, try to read it from the ibft firmware.

12 years agodracut-lib.sh: add killproc()
Harald Hoyer [Wed, 28 Sep 2011 14:02:59 +0000 (16:02 +0200)]
dracut-lib.sh: add killproc()

12 years ago50plymouth/plymouth-pretrigger.sh: respect primary console setting
Harald Hoyer [Wed, 9 Nov 2011 09:18:58 +0000 (10:18 +0100)]
50plymouth/plymouth-pretrigger.sh: respect primary console setting

The kernel's primary console device is determined by the last "console="
argument on the kernel command line. This setting should be respected by
dracut-generated initial RAM disks.

Steps to Reproduce:
(Easiest using a KVM VM, virt-manager and "virsh console")
1. Boot with a kernel command line ending in
 console=tty0 console=ttyS0,115200
2. Observe both tty0 and ttyS0.

The output of init scripts is sent to ttyS0, as the final "console="
argument determines the primary console device as per
Documentation/serial-console.txt in the kernel sources.

https://bugzilla.redhat.com/show_bug.cgi?id=752073

12 years ago99base/init: avoid using cp -t
Daniel Drake [Tue, 8 Nov 2011 21:11:40 +0000 (21:11 +0000)]
99base/init: avoid using cp -t

-t is not supported by busybox's cp.

Signed-off-by: Daniel Drake <dsd@laptop.org>
12 years agodracut.spec: remove xen module
Harald Hoyer [Wed, 9 Nov 2011 08:55:16 +0000 (09:55 +0100)]
dracut.spec: remove xen module

12 years agoMakefile, git2spec.pl: use temporary builddir
Harald Hoyer [Wed, 9 Nov 2011 08:51:15 +0000 (09:51 +0100)]
Makefile, git2spec.pl: use temporary builddir

to build the dracut rpm, use a temporary builddir

12 years ago99fs-lib/module-setup.sh: only include fsck helper needed for hostonly
Harald Hoyer [Wed, 9 Nov 2011 08:04:59 +0000 (09:04 +0100)]
99fs-lib/module-setup.sh: only include fsck helper needed for hostonly

for the host-only initramfs, check all relevant disks for the
filesystem type and only include those fsck tools, which are needed.

12 years ago99base/init, 98usrmount: mount /usr if found in /sysroot/etc/fstab
Harald Hoyer [Mon, 7 Nov 2011 07:52:03 +0000 (08:52 +0100)]
99base/init, 98usrmount: mount /usr if found in /sysroot/etc/fstab

We cannot boot correctly without /usr, because of
http://www.freedesktop.org/wiki/Software/systemd/separate-usr-is-broken
so, we mount /usr readonly.

12 years agoadd wait_for_dev() and wait_for_mount()
Harald Hoyer [Mon, 7 Nov 2011 07:46:36 +0000 (08:46 +0100)]
add wait_for_dev() and wait_for_mount()

12 years agofs-lib: install /etc/e2fsck.conf, if e2fsck is installed
Harald Hoyer [Wed, 2 Nov 2011 08:09:48 +0000 (09:09 +0100)]
fs-lib: install /etc/e2fsck.conf, if e2fsck is installed

12 years ago90kernel-modules/module-setup.sh: include usb mass storage drivers
Harald Hoyer [Mon, 7 Nov 2011 08:37:55 +0000 (09:37 +0100)]
90kernel-modules/module-setup.sh: include usb mass storage drivers

12 years agoinst_script should call inst_simple (#750603)
Brian C. Lane [Tue, 1 Nov 2011 19:13:57 +0000 (12:13 -0700)]
inst_script should call inst_simple (#750603)

inst_script checks for a shebang, if it doesn't exist it exits.
If it does it should not be calling inst_binary, it should call
inst_simple like it used to.

12 years agoapply-live-updates.sh: copy without glob
Will Woods [Thu, 20 Oct 2011 17:57:23 +0000 (13:57 -0400)]
apply-live-updates.sh: copy without glob

cp $SRC/* $DEST will skip dotfiles in $SRC.
( cd $SRC; cp -a -t $DEST . ) will copy everything.

12 years agoremove the xen dracut module
Paolo Bonzini [Thu, 20 Oct 2011 17:15:06 +0000 (19:15 +0200)]
remove the xen dracut module

The Xen module is unnecessary and it has been for a while.
Most Xen systems will not be using the module, even now, because
xen-detect is not installed by default on most Xen systems, and
dracut uses xen-detect to decide whether to include the module.

It also has some problems:

1) it does not try loading xen_platform_pci;

2) it loads modules unnecessarily; modules.alias is where all Xen support
should reside.  Assuming xenbus_probe_frontend and xen_platform_pci
are loaded so that Xen devices are probed, other modules are picked up
automatically thanks to aliases such as

    alias xen:vbd xen_blkfront

3) Even not-so-recent kernels (say 2.6.32) require the xen_platform_pci
and xenbus_probe_frontend modules even for non-paravirtualized guests.
60xen/module-setup.sh picks the module only for PV guests.

So, just require xenbus_probe_frontend to be builtin, and also
xen_platform_pci for fully-virtualized guests, and remove the module.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
12 years agodracut.spec: require new util-linux version with new switch_root
Harald Hoyer [Fri, 21 Oct 2011 10:35:34 +0000 (12:35 +0200)]
dracut.spec: require new util-linux version with new switch_root

new switch_root knows how to handle "/run"

12 years agodracut.spec: apply patches with git
Harald Hoyer [Fri, 21 Oct 2011 10:35:07 +0000 (12:35 +0200)]
dracut.spec: apply patches with git

This will take all renames and mode changes.

12 years agodracut-functions:inst_script() call inst_binary() with found bin
Harald Hoyer [Fri, 21 Oct 2011 10:33:05 +0000 (12:33 +0200)]
dracut-functions:inst_script() call inst_binary() with found bin

call inst_binary() with the already found executable to prevent
searching for the binary again

12 years agodracut-functions: do not install files from current directory
Harald Hoyer [Fri, 21 Oct 2011 08:09:55 +0000 (10:09 +0200)]
dracut-functions: do not install files from current directory

Protect against relative pathnames without a slash for all inst_*()
functions.

12 years ago90livenet: do not install by default
Harald Hoyer [Fri, 21 Oct 2011 08:15:09 +0000 (10:15 +0200)]
90livenet: do not install by default

12 years ago90dmsquash-live: do not install by default
Harald Hoyer [Fri, 21 Oct 2011 08:14:49 +0000 (10:14 +0200)]
90dmsquash-live: do not install by default

12 years ago90livenet/module-setup.sh: silence check for wget
Harald Hoyer [Fri, 21 Oct 2011 07:35:38 +0000 (09:35 +0200)]
90livenet/module-setup.sh: silence check for wget

12 years agodracut-functions: inst_dir() handle relative symlinks
Harald Hoyer [Thu, 20 Oct 2011 11:04:30 +0000 (13:04 +0200)]
dracut-functions: inst_dir() handle relative symlinks

12 years agodracut-logger: re-set debugging
Harald Hoyer [Thu, 20 Oct 2011 11:04:10 +0000 (13:04 +0200)]
dracut-logger: re-set debugging

12 years ago90livenet: check() for wget
Harald Hoyer [Thu, 20 Oct 2011 10:33:33 +0000 (12:33 +0200)]
90livenet: check() for wget

12 years agominor changes
Michal Soltys [Tue, 18 Oct 2011 18:33:19 +0000 (20:33 +0200)]
minor changes

- in 10i18n - do stty -iutf8 on non-utf8 consoles, for consistency with
  iutf8 on utf8 ones
- vim modeline in xml file

Signed-off-by: Michal Soltys <soltys@ziu.info>
12 years ago99base/dracut-lib.sh: add /etc/cmdline/*.conf parsing
Harald Hoyer [Thu, 20 Oct 2011 09:14:10 +0000 (11:14 +0200)]
99base/dracut-lib.sh: add /etc/cmdline/*.conf parsing

modules and overlay images can set default kernel command line
parameters in /etc/cmdline/*.conf in the initramfs.

12 years ago99base/init: fix check for in kernel polling
Harald Hoyer [Thu, 20 Oct 2011 09:13:34 +0000 (11:13 +0200)]
99base/init: fix check for in kernel polling

12 years ago*/module-setup.sh: turn off debugging for module search
Harald Hoyer [Thu, 20 Oct 2011 10:13:50 +0000 (12:13 +0200)]
*/module-setup.sh: turn off debugging for module search

turn off debugging for module search, because it clutters the debug log

12 years ago95udev-rules/module-setup.sh: correctly create udevd symlink
Harald Hoyer [Thu, 20 Oct 2011 10:13:25 +0000 (12:13 +0200)]
95udev-rules/module-setup.sh: correctly create udevd symlink

12 years agonetroot: do not die, if arping failed
Harald Hoyer [Wed, 19 Oct 2011 12:24:07 +0000 (14:24 +0200)]
netroot: do not die, if arping failed

12 years agocheck root candidates more carefully
Michal Soltys [Mon, 17 Oct 2011 21:36:33 +0000 (23:36 +0200)]
check root candidates more carefully

This is from the following thread:

http://thread.gmane.org/gmane.linux.raid/35753/focus=35795

Additional tests + more specific info.

Signed-off-by: Michal Soltys <soltys@ziu.info>
[harald@redhat.com: usable_root(): relaxed check for root]

12 years agodmsquash-live: fix log message about root/liveroot
Will Woods [Thu, 13 Oct 2011 02:49:35 +0000 (22:49 -0400)]
dmsquash-live: fix log message about root/liveroot

12 years agodracut-lib.sh: fix dropped backslashes in CMDLINE
Will Woods [Thu, 13 Oct 2011 02:48:08 +0000 (22:48 -0400)]
dracut-lib.sh: fix dropped backslashes in CMDLINE

The "read" shell builtin consumes backslashes, which is a problem if
your root device is something like "LABEL=Fedora\x2016".

Using "read -r" tells the shell to leave backslashes alone.

12 years agodracut: remove duplicate options
Michal Soltys [Mon, 17 Oct 2011 21:01:49 +0000 (23:01 +0200)]
dracut: remove duplicate options

'-I' and '--fwdir' were both read_arg and push_arg, and the latter has
priority.

Signed-off-by: Michal Soltys <soltys@ziu.info>
12 years agoexplicitly verify bash version
Michal Soltys [Mon, 10 Oct 2011 21:58:05 +0000 (23:58 +0200)]
explicitly verify bash version

A simple check in main dracut script.

Signed-off-by: Michal Soltys <soltys@ziu.info>
12 years agobash3 compat patch
Michal Soltys [Mon, 10 Oct 2011 21:58:04 +0000 (23:58 +0200)]
bash3 compat patch

This patch replaces:

- {var}>... redirections with functionally identical eval construct +
  explicit FDs
- ^^ and ,, case modifiers with temporary shopt

This allows us to lower minimum required bash version
to at least 3.1 (with current code).

Signed-off-by: Michal Soltys <soltys@ziu.info>
12 years agodracut.conf.5.xml: tag mismatch fix
Dave Young [Tue, 11 Oct 2011 03:26:54 +0000 (11:26 +0800)]
dracut.conf.5.xml: tag mismatch fix

build failed with:
xsltproc -o dracut.conf.5 -nonet http://docbook.sourceforge.net/release/xsl/current/manpages/docbook.xsl dracut.conf.5.xml
dracut.conf.5.xml:169: parser error : Opening and ending tag mismatch: para line 168 and listitem
        </listitem>

Fix it by change to </para> at the end

Signed-off-by: Dave Young <dyoung@redhat.com>
12 years ago95udev-rules/module-setup.s: fixed symlink for udevd to initdir
Harald Hoyer [Mon, 10 Oct 2011 18:17:16 +0000 (20:17 +0200)]
95udev-rules/module-setup.s: fixed symlink for udevd to initdir

12 years agodracut.*.xml: s/exisiting/existing/g
Harald Hoyer [Mon, 10 Oct 2011 09:55:17 +0000 (11:55 +0200)]
dracut.*.xml: s/exisiting/existing/g

12 years agodracut-functions: conv/normalize minor corrections
Michal Soltys [Fri, 7 Oct 2011 22:20:50 +0000 (00:20 +0200)]
dracut-functions: conv/normalize minor corrections

mostly with reference to earlier commit:

- bash doesn't need unsetting locals
- make normalize_path() a bit faster, also make sure we remove all
  trailing slashes
- normalize paths before tests

Signed-off-by: Michal Soltys <soltys@ziu.info>
12 years agomanuals: add info about fs-lib (fsck) configuration
Michal Soltys [Fri, 7 Oct 2011 20:23:52 +0000 (22:23 +0200)]
manuals: add info about fs-lib (fsck) configuration

in dracut.8.xml & dracut.conf.5.xml

Signed-off-by: Michal Soltys <soltys@ziu.info>
12 years agofs-lib: add ability to choose fsck tools
Michal Soltys [Fri, 7 Oct 2011 20:23:51 +0000 (22:23 +0200)]
fs-lib: add ability to choose fsck tools

in dracut.conf:

fscks="<tools>"
nofscks="yes"

and similary on command line:

--fscks [LIST] (in addition to conf's, if defined there)
--nofscks

Signed-off-by: Michal Soltys <soltys@ziu.info>
12 years agodracut.8: add missing lvmconf info
Michal Soltys [Fri, 7 Oct 2011 20:23:50 +0000 (22:23 +0200)]
dracut.8: add missing lvmconf info

Signed-off-by: Michal Soltys <soltys@ziu.info>
12 years agoconvert_abs_rel() fixups
Michal Soltys [Fri, 7 Oct 2011 20:23:49 +0000 (22:23 +0200)]
convert_abs_rel() fixups

- IFS was not preserved, and modified value could leak to outside functions

- the '.' relative path should be returned for arguments such as /x/y/z
  /x/y - but not for $1 == $2 ones

- $1 == $2 is self-looping link, so it returns final component of its
  name

Signed-off-by: Michal Soltys <soltys@ziu.info>
12 years agodracut: check mktemp return value
Dave Young [Mon, 10 Oct 2011 09:41:14 +0000 (11:41 +0200)]
dracut: check mktemp return value

in slackware the default mktemp is not from coreutils.
A simply make in test directory mangled my rootfs due
to initdir is blank

Also mktemp could failed with other reason like ENOSPC or EPERM

12 years agoupdate the documentation of '--[no]prefix'
WANG Cong [Fri, 7 Oct 2011 07:44:10 +0000 (15:44 +0800)]
update the documentation of '--[no]prefix'

In

commit fd786adcf515d9d3ee77eb29fa4c6b60020c7209
Author: Harald Hoyer <harald@redhat.com>
Date:   Wed Apr 20 16:47:40 2011 +0200

    dracut: make prefix configurable

Harald changed the meaning of --prefix/--noprefix, but
forgot to update their documentation. This patch
fixes that.

Signed-off-by: WANG Cong <xiyou.wangcong@gmail.com>
12 years ago95udev-rules/module-setup.sh: also search in /lib/udev and /usr/lib/udev
Harald Hoyer [Thu, 6 Oct 2011 08:25:08 +0000 (10:25 +0200)]
95udev-rules/module-setup.sh: also search in /lib/udev and /usr/lib/udev

12 years ago90mdraid/module-setup.sh: fixed sed arguments
Harald Hoyer [Tue, 4 Oct 2011 11:30:35 +0000 (13:30 +0200)]
90mdraid/module-setup.sh: fixed sed arguments

12 years ago90mdraid/mdraid_start.sh: fix path to md sysfs
Harald Hoyer [Tue, 4 Oct 2011 11:06:33 +0000 (13:06 +0200)]
90mdraid/mdraid_start.sh: fix path to md sysfs

12 years agoorder mdadm and lvm timeout operations
Harald Hoyer [Tue, 4 Oct 2011 11:03:45 +0000 (13:03 +0200)]
order mdadm and lvm timeout operations

12 years agodracut [PATCH]es: parallelize block_module filter and net_module_filter
John Reiser [Fri, 23 Sep 2011 16:17:13 +0000 (09:17 -0700)]
dracut [PATCH]es: parallelize block_module filter and net_module_filter

Filtering modules requires enough work that instmods() in the
next pipeline stage was rarely busy.  Parallelize the two
filters which do the most work.  Also fix a filename-vs-contents
mistake in net_module_filter.

--
John Reiser, jreiser@BitWagon.com

>From f4533a2ceca52c443ddebec01eeaa35d51c39c1b Mon Sep 17 00:00:00 2001
From: John Reiser <jreiser@BitWagon.com>
Date: Tue, 13 Sep 2011 17:41:43 -0700
Subject: [PATCH 1/3] Parallelize block_module_filter

12 years agodracut [PATCH] Parameter expansion occurs before command evaluation.
John Reiser [Fri, 23 Sep 2011 15:02:23 +0000 (08:02 -0700)]
dracut [PATCH] Parameter expansion occurs before command evaluation.

Bash shell expands all ${parameter} before evaluating a command.
For multiple declarations and assignments within the same 'local' command,
then new variables or new values that appear towards the left
do not affect parameter expansion towards the right.

--
John Reiser, jreiser@BitWagon.com

>From 507ad6f66fc66f868a9e5fdd3806e012c4022baa Mon Sep 17 00:00:00 2001
From: John Reiser <jreiser@BitWagon.com>
Date: Fri, 23 Sep 2011 07:37:43 -0700
Subject: [PATCH] Parameter expansion occurs before command evaluation.

${parameter} on the right is expanded before evaluating "local var=value"
on the left.

12 years ago90mdraid: adjust /dev/md loops
Michal Soltys [Thu, 22 Sep 2011 15:16:39 +0000 (17:16 +0200)]
90mdraid: adjust /dev/md loops

Include '_' to also handle old partitionable arrays (pre-2.6.28).

Signed-off-by: Michal Soltys <soltys@ziu.info>
12 years ago90mdraid: more thorough 64-md-raid.rules edit
Michal Soltys [Thu, 22 Sep 2011 15:23:21 +0000 (17:23 +0200)]
90mdraid: more thorough 64-md-raid.rules edit

Catch more variations of incremental assembly.

Signed-off-by: Michal Soltys <soltys@ziu.info>
12 years ago90mdraid: check precisely for supported contaiers
Michal Soltys [Thu, 22 Sep 2011 15:09:56 +0000 (17:09 +0200)]
90mdraid: check precisely for supported contaiers

ID_FS_TYPE can be much more than just ddf/imsm/linux raid member, so
do the proper checks.

This reverts certain changes from:
cf5891424ef026eede69606a918dadf5560095fd

Signed-off-by: Michal Soltys <soltys@ziu.info>
12 years agodracut-functions: inst_rules() add missing ""
Harald Hoyer [Thu, 22 Sep 2011 14:14:38 +0000 (16:14 +0200)]
dracut-functions: inst_rules() add missing ""

12 years ago99fs-lib: export FSTAB_FILE before fsck call
Michal Soltys [Mon, 5 Sep 2011 23:22:15 +0000 (01:22 +0200)]
99fs-lib: export FSTAB_FILE before fsck call

Signed-off-by: Michal Soltys <soltys@ziu.info>
12 years agodracut: create /dev besides /proc, /sys and so
Amadeusz Żołnowski [Sun, 4 Sep 2011 14:38:35 +0000 (16:38 +0200)]
dracut: create /dev besides /proc, /sys and so

How it worked without it?  The issue only manifests itself with
initramfs integrated into kernel.

12 years agodracut-logger: bail out early, if we don't have to log anything
Harald Hoyer [Thu, 22 Sep 2011 13:49:25 +0000 (15:49 +0200)]
dracut-logger: bail out early, if we don't have to log anything

12 years agostr_replace() fix
Michal Soltys [Thu, 15 Sep 2011 06:45:57 +0000 (08:45 +0200)]
str_replace() fix

Whitespace removal in:

out="${out}${chop# }$r"

will damage certain strings, for example the following call:

str_replace ' aax aaxaa' x y

would return 'aayaayaa' instead of ' aay aayaa'.

Signed-off-by: Michal Soltys <soltys@ziu.info>
12 years agodracut-functions: inst_rules() do not check std dirs for abs path
Harald Hoyer [Thu, 22 Sep 2011 13:43:34 +0000 (15:43 +0200)]
dracut-functions: inst_rules() do not check std dirs for abs path

if an absolute path is given, we should not check the standard udev rule
directories.

12 years ago99fs-lib/fs-lib.sh: fsck btrfs via mounting like xfs
Harald Hoyer [Thu, 22 Sep 2011 13:12:14 +0000 (15:12 +0200)]
99fs-lib/fs-lib.sh: fsck btrfs via mounting like xfs

btrfsck is only for manual repairing your filesystem

12 years agoTEST-40-NBD: relaxed check on ext3 filesystem options
Harald Hoyer [Thu, 22 Sep 2011 10:56:31 +0000 (12:56 +0200)]
TEST-40-NBD: relaxed check on ext3 filesystem options

onerror=continue does not seem to be displayed for new kernels

12 years ago90mdraid: fix/adjust 65-md* rules and related scripts
Michal Soltys [Mon, 5 Sep 2011 22:17:28 +0000 (00:17 +0200)]
90mdraid: fix/adjust 65-md* rules and related scripts

Reworked the flow of the rules file a bit, removed redundant tests, also
should be easier to follow. It's much shorter now as well, a bit more
similar to 90lvm script - both revolve around same concepts after all.

There's no reason to treat conf-assembled arrays differently from
incremental ones. Once we hit timeout in init's udev loop, we can use
common script (mdraid_start.sh) to try force inactive arrays
into degraded mode.

md-finished.sh was kind-of out of place - it didn't really wait for any
particular device(s) to show up, just watched if onetime mdadm scripts
are still in place. Furthermore, after moving mdraid_start to --timeout
initqueue, it didn't really have too much to watch at all, besides
mdadm_auto (and that served no purpose, as we do wait for concrete
devices).

Either way, with stock 64-md fixes, current version of 65-md*.rules does
the following:

- limits assembly to certain uuids, if specified
- watch for no ddf/imsm
- if mdadm.conf => setup onetime -As script, without forced --run option
- if !mdadm.conf => incrementally assemble
- for both cases, setup timeout script, run-forcing arrays as a last resort

Signed-off-by: Michal Soltys <soltys@ziu.info>
12 years ago90(md|dm)raid: recognize ddf container
Michal Soltys [Mon, 5 Sep 2011 22:17:24 +0000 (00:17 +0200)]
90(md|dm)raid: recognize ddf container

This patch adds ddf support, consistently with imsm/isw options.

Signed-off-by: Michal Soltys <soltys@ziu.info>
12 years ago90mdraid: fix/adjust force-run script
Michal Soltys [Mon, 5 Sep 2011 22:17:27 +0000 (00:17 +0200)]
90mdraid: fix/adjust force-run script

1) mdadm -As --auto=yes --run 2>&1 | vinfo (removed)

Currently such auto assembly will not complete or force-run partially
assembled arrays. It might assemble "concurrent" separate array and
force-run it, if possible (though the chances of suddenly showing
missing components in this scenario - a script run after udev timeout -
are pretty thin). See [1] for details. Also see #3 below.

2) mdadm -Is --run 2>&1 (removed)

This will only force-run native arrays - arrays in containers will not
be affected. See [1] for details. Also see #3 below.

3) mdadm -R run loop (implicitly handles #1 & #2)

This loop does everywthing that #1 & #2 are expected to do. Thus, the
above invocations are simply redundant and this is the most safe and
flexible option.

Also, it shouldn't be necessary to go under md/ directory, as those are
just symlinks to /dev/md[0-9]*.

Certain checks were changed to strict ones (array state, degraded state)
instead of relying on env tricks.

'cat' was added explicitly to installed programs (it has been used
implicitly in shutdown script either way)

4) mdmon bug

See [1] for details as well. In short - force-run arrays in containers
will not have mdmon started, so we do that manually.

5) stop/run queue magic

Also removed. mdadm -R will only cause change events to the array
itself, and they should not be an issue.

[1] http://article.gmane.org/gmane.linux.raid/35133

Signed-off-by: Michal Soltys <soltys@ziu.info>
12 years ago90mdraid: fix/adjust mdraid-cleanup
Michal Soltys [Mon, 5 Sep 2011 22:17:26 +0000 (00:17 +0200)]
90mdraid: fix/adjust mdraid-cleanup

Stop both arrays (first pass) and containers (second pass).
Loop only over /dev/md[0-9]*
Call cleanup script only once, make sure it's after crypt cleanup.

Signed-off-by: Michal Soltys <soltys@ziu.info>
12 years ago90mdraid: containers are not runnable
Michal Soltys [Mon, 5 Sep 2011 22:17:25 +0000 (00:17 +0200)]
90mdraid: containers are not runnable

Remove whole "start a container logic".

Containers once assembled, always remain in 'inactive' state.
Any attempt to run a container with mdadm -IR is a no-op, and any
attempt with just mdadm -R ends with an error.

Signed-off-by: Michal Soltys <soltys@ziu.info>
12 years ago90mdraid: adjust stock mdadm udev rules
Michal Soltys [Mon, 5 Sep 2011 22:17:23 +0000 (00:17 +0200)]
90mdraid: adjust stock mdadm udev rules

Currently shipped mdadm rules incrementally assemble all imsm and native
raids, and do so unconditionally. This causes few issues:

- fine-grained controls in 65-md* are shadowed - for example,
  mdadm.conf's presence tests or uuid checks
- 90dmraid might also conflict with 90mdraid, if user prefers the former
  to handle containers
- possibly other subtle issues

This patch adjusts the behaviour.

Signed-off-by: Michal Soltys <soltys@ziu.info>
12 years agomount securityfs in a seperate dracut module
Harald Hoyer [Tue, 20 Sep 2011 09:16:53 +0000 (11:16 +0200)]
mount securityfs in a seperate dracut module

12 years agomount securityfs in a seperate dracut module
Harald Hoyer [Tue, 20 Sep 2011 09:16:53 +0000 (11:16 +0200)]
mount securityfs in a seperate dracut module

12 years ago99base/init: mount securityfs with source "securityfs" instead of dest
Harald Hoyer [Tue, 20 Sep 2011 08:10:29 +0000 (10:10 +0200)]
99base/init: mount securityfs with source "securityfs" instead of dest

12 years agoDo not use /run/udev/rules.d for udev rules
Harald Hoyer [Mon, 19 Sep 2011 10:21:51 +0000 (12:21 +0200)]
Do not use /run/udev/rules.d for udev rules

for rules, which should not be called in the real root.

Stale rules can cause this:
https://bugzilla.redhat.com/show_bug.cgi?id=734096

12 years ago99base/init: remove /dev/root helper symlink
Harald Hoyer [Mon, 19 Sep 2011 10:20:55 +0000 (12:20 +0200)]
99base/init: remove /dev/root helper symlink

Any tool relying on /dev/root has to be fixed.

12 years ago90dmsquash-live: do not symlink to /dev/live
Harald Hoyer [Mon, 19 Sep 2011 10:20:11 +0000 (12:20 +0200)]
90dmsquash-live: do not symlink to /dev/live

/dev/live should not be used anyway

12 years agodmsquash-live-root: load filesystem modules before mounting loop images
Harald Hoyer [Fri, 2 Sep 2011 17:16:17 +0000 (19:16 +0200)]
dmsquash-live-root: load filesystem modules before mounting loop images

might prevent https://bugzilla.redhat.com/show_bug.cgi?id=735199

12 years agodmsquash-live-root: use blkid to determine fstype of images
Harald Hoyer [Fri, 2 Sep 2011 17:01:16 +0000 (19:01 +0200)]
dmsquash-live-root: use blkid to determine fstype of images

prevents:
dracut: FATAL: cannot mount live image (unknown filesystem type)
https://bugzilla.redhat.com/show_bug.cgi?id=735199

12 years ago99base/init: removed cdrom polling reset code
Harald Hoyer [Fri, 2 Sep 2011 07:01:05 +0000 (09:01 +0200)]
99base/init: removed cdrom polling reset code

This is done globally now.

12 years agodracut: cp with sparse
Harald Hoyer [Fri, 2 Sep 2011 07:01:47 +0000 (09:01 +0200)]
dracut: cp with sparse

12 years ago99base/init: do not fail, when importing the original kernel environment
Harald Hoyer [Thu, 1 Sep 2011 14:01:21 +0000 (16:01 +0200)]
99base/init: do not fail, when importing the original kernel environment

12 years agoinstmods: factor out egrep of "FATAL: Module .* not found"
John Reiser [Mon, 29 Aug 2011 23:03:35 +0000 (16:03 -0700)]
instmods: factor out egrep of "FATAL: Module .* not found"

12 years agoinstmods: sanity for _mpargs
John Reiser [Mon, 29 Aug 2011 21:46:25 +0000 (14:46 -0700)]
instmods: sanity for _mpargs

12 years agoinstmods: get filenames from stdin if no args; use it
John Reiser [Mon, 29 Aug 2011 20:40:21 +0000 (13:40 -0700)]
instmods: get filenames from stdin if no args; use it

Use bash  "[[ string =~ pattern ]]"  instead of "egrep -q".
Replace control-dominated serial fondling
for var in $(proc1); do proc2 var; done
with data-dominated parallel pipeline
proc1  |  while read var; do proc2 var; done
Together this is a large savings.

[harald@redhat.com: fixed network kernel module filter]

12 years agoinstall_kmod_with_fw: make fast case faster
John Reiser [Sun, 28 Aug 2011 20:24:58 +0000 (13:24 -0700)]
install_kmod_with_fw: make fast case faster

12 years agofilter_kernel_modules is a specialized filter_kernel_modules_by_path
John Reiser [Sat, 27 Aug 2011 21:43:49 +0000 (14:43 -0700)]
filter_kernel_modules is a specialized filter_kernel_modules_by_path

12 years agoinst_simple, inst_dir: make fast case faster
John Reiser [Fri, 26 Aug 2011 20:01:33 +0000 (13:01 -0700)]
inst_simple, inst_dir: make fast case faster

This small stuff saves 1.7% per dropped statement during "dracut --profile".
Fixing the comment about /lib -> lib64 is REQUIRED!

12 years ago95udev-rules: add input_id
Harald Hoyer [Wed, 31 Aug 2011 14:48:20 +0000 (16:48 +0200)]
95udev-rules: add input_id

12 years agodracut-functions: hmac checksum files can be symlinks, too
Harald Hoyer [Wed, 31 Aug 2011 13:22:09 +0000 (15:22 +0200)]
dracut-functions: hmac checksum files can be symlinks, too

use inst() instead of inst_simple() to install the hmac files

12 years ago99base/init: move switch_root breakpoint to a later point in the script
Harald Hoyer [Tue, 30 Aug 2011 14:23:17 +0000 (16:23 +0200)]
99base/init: move switch_root breakpoint to a later point in the script

12 years ago99base/init: save and restore environment given from the kernel
Harald Hoyer [Tue, 30 Aug 2011 14:22:46 +0000 (16:22 +0200)]
99base/init: save and restore environment given from the kernel

12 years ago90crypt/parse-crypt.sh: also accept the beginning of the LUKS UUID
Harald Hoyer [Tue, 30 Aug 2011 12:43:57 +0000 (14:43 +0200)]
90crypt/parse-crypt.sh: also accept the beginning of the LUKS UUID

2e0c003435bbc0751cdf7466c0faebe7bfc7445b introduced a too strict test
for LUKS UUIDs

12 years agobuild initramfs: unclear _mpargs in instmods()
John Reiser [Mon, 29 Aug 2011 21:42:15 +0000 (14:42 -0700)]
build initramfs: unclear _mpargs in instmods()

The local variable _mpargs in function instmods() in file dracut-functions
looks peculiar.  The documentation is non-existent, but still ...

First, $_mpargs is not passed to modprobe via for_each_kmod_dep.
This is strange because my guess is that "_mpargs" means
"extra arguments for modprobe".

Second, the leading "--" will be lopped when a leading pathname
is stripped via
_mod=${_mod##*/}
It seems to me that a leading "--" should inhibit modification.

Here's the corresponding patch to current HEAD (from dracut-013.)

12 years ago99base/init: only poll cdroms, if the kernel does support autopolling
Harald Hoyer [Mon, 29 Aug 2011 17:12:12 +0000 (19:12 +0200)]
99base/init: only poll cdroms, if the kernel does support autopolling

12 years ago50plymouth: add plymouth.enable kernel command line option
Harald Hoyer [Mon, 29 Aug 2011 16:30:26 +0000 (18:30 +0200)]
50plymouth: add plymouth.enable kernel command line option

12 years agodracut.spec: fixed rhel/fedora version checks
Harald Hoyer [Mon, 29 Aug 2011 11:11:49 +0000 (13:11 +0200)]
dracut.spec: fixed rhel/fedora version checks

12 years ago90mdraid/65-md-incremental-imsm.rules: incremental run to settled
Harald Hoyer [Wed, 24 Aug 2011 14:35:33 +0000 (16:35 +0200)]
90mdraid/65-md-incremental-imsm.rules: incremental run to settled

move incremental run to settled queue again

https://bugzilla.redhat.com/show_bug.cgi?id=732967

12 years agodracut-functions: fix inst_dir() for non-absolute dirs
Harald Hoyer [Tue, 23 Aug 2011 10:50:03 +0000 (12:50 +0200)]
dracut-functions: fix inst_dir() for non-absolute dirs

12 years agoluks key on ext dev - wait for luks
Przemysław Rudy [Mon, 22 Aug 2011 09:27:00 +0000 (11:27 +0200)]
luks key on ext dev - wait for luks

This really waits for the luks mapper device, so luksOpen can do it job