chiark / gitweb /
elogind.git
9 years agobootchart: clean up sysfd and proc handling
Daniel Mack [Thu, 2 Apr 2015 12:15:33 +0000 (14:15 +0200)]
bootchart: clean up sysfd and proc handling

Retrieve the handle to procfs in main(), and pass it functions
that need it. Kill the global variables.

Also, refactor lots of code in svg_title(). There's no need to access any
global variables from there either, and we really should return proper
errors from there as well.

9 years agobootchart: clean up control flow logic
Daniel Mack [Thu, 2 Apr 2015 11:43:18 +0000 (13:43 +0200)]
bootchart: clean up control flow logic

Don't blindly exit() from random functions, but return a proper error
and upchain error conditions.

squash! bootchart: clean up control flow logic

When pread() returns "0", it's a read failure, so don't make the caller think
log_sample() was successful, return meaningful error code instead of 0.

9 years agobootchart: switch to log_* helpers
Daniel Mack [Thu, 2 Apr 2015 11:24:30 +0000 (13:24 +0200)]
bootchart: switch to log_* helpers

Let the helper functions take care of the string message output.

9 years agoupdate TODO
Lennart Poettering [Fri, 3 Apr 2015 12:34:59 +0000 (14:34 +0200)]
update TODO

9 years agounits: explicitly require /var, /tmp and /var/tmp to be mounted before basic.target
Lennart Poettering [Fri, 3 Apr 2015 12:32:32 +0000 (14:32 +0200)]
units: explicitly require /var, /tmp and /var/tmp to be mounted before basic.target

We support /var, /tmp and /var/tmp on NFS. NFS shares however are by
default ordered only before remote-fs.target which is a late-boot
service. /var, /tmp, /var/tmp need to be around earlier though, hence
explicitly order them before basic.target.

Note that this change simply makes explicit what was implicit before,
since many early-boot services pulled in parts of /var anyway early.

9 years agounits: explicitly order systemd-user-sessions.service after nss-user-lookup.target
Lennart Poettering [Fri, 3 Apr 2015 12:31:35 +0000 (14:31 +0200)]
units: explicitly order systemd-user-sessions.service after nss-user-lookup.target

We should not allow logins before NIS/LDAP users are available.

9 years agounits: move After=systemd-hwdb-update.service dependency from udev to udev-trigger
Lennart Poettering [Fri, 3 Apr 2015 12:27:16 +0000 (14:27 +0200)]
units: move After=systemd-hwdb-update.service dependency from udev to udev-trigger

Let's move the hwdb regeneration a bit later. Given that hwdb is
non-essential it should be OK to allow udev to run without it until we
do the full trigger.

http://lists.freedesktop.org/archives/systemd-devel/2015-April/030074.html

9 years agoCODING_STYLE: mention that dup() should not be used
Lennart Poettering [Fri, 3 Apr 2015 12:26:22 +0000 (14:26 +0200)]
CODING_STYLE: mention that dup() should not be used

9 years agotimedatectl: many fixes
Lennart Poettering [Thu, 2 Apr 2015 10:15:53 +0000 (12:15 +0200)]
timedatectl: many fixes

- print runtime warnings with log_warning()

- save and restore $TZ properly

- Get rid of exit() pseudo error handling

- Using time() is OK when connecting to a local container or when
  showing data about local host, but certainly not for remote hosts.

9 years agoman: don't mention "journalctl /dev/sda"
Lennart Poettering [Thu, 2 Apr 2015 10:14:57 +0000 (12:14 +0200)]
man: don't mention "journalctl /dev/sda"

It never worked, and nobody ever worked on it, hence don't mention it.

9 years agoupdate TODO
Lennart Poettering [Thu, 2 Apr 2015 10:14:51 +0000 (12:14 +0200)]
update TODO

9 years agoCODING_STYLE: clarify that exit() is never OK to call
Lennart Poettering [Thu, 2 Apr 2015 10:14:43 +0000 (12:14 +0200)]
CODING_STYLE: clarify that exit() is never OK to call

9 years agolibudev: device - replace by a thin wrapper around sd-device
Tom Gundersen [Wed, 1 Apr 2015 11:55:20 +0000 (13:55 +0200)]
libudev: device - replace by a thin wrapper around sd-device

9 years agofsck: port from libudev to sd-device
Tom Gundersen [Wed, 1 Apr 2015 12:33:02 +0000 (14:33 +0200)]
fsck: port from libudev to sd-device

9 years agocryptsetup: port from libudev to sd-device
Tom Gundersen [Wed, 1 Apr 2015 12:26:47 +0000 (14:26 +0200)]
cryptsetup: port from libudev to sd-device

9 years agonetworkctl: port from libudev to sd-device
Tom Gundersen [Fri, 5 Dec 2014 00:16:05 +0000 (01:16 +0100)]
networkctl: port from libudev to sd-device

9 years agolibsystemd: add sd-device library
Tom Gundersen [Wed, 1 Apr 2015 11:50:31 +0000 (13:50 +0200)]
libsystemd: add sd-device library

This provides equivalent functionality to libudev-device, but in the
systemd style. The public API only caters to creating sd_device objects
from for devices that already exist in /sys, there is no support for
listening for monitoring events or creating devices received over
the udev netlink protocol.

The private API contains the necessary functionality to make sd-device
a drop-in replacement for libudev-device, but which we would not
otherwise want to export.

9 years agoudev: net_id - improve comments
Tom Gundersen [Wed, 1 Apr 2015 21:34:19 +0000 (23:34 +0200)]
udev: net_id - improve comments

The dev_port concept is a bit confusing, expand on the comment a bit.

9 years agoudev: net_id - support multi-port enpo* device names
Tom Gundersen [Wed, 1 Apr 2015 14:51:02 +0000 (16:51 +0200)]
udev: net_id - support multi-port enpo* device names

I'd argue that having firmware labels for such devices makes
no sense, but they exist, so make sure we handle them as best
as we can.

9 years agoudev: net_id - fix copy-paste error
Tom Gundersen [Wed, 1 Apr 2015 14:41:41 +0000 (16:41 +0200)]
udev: net_id - fix copy-paste error

In case pci_slot overflows we were truncating pci_path instead.

9 years agolibudev: monitor - improve logging
Tom Gundersen [Wed, 1 Apr 2015 11:55:52 +0000 (13:55 +0200)]
libudev: monitor - improve logging

9 years agoshared: set - make argument to set_remove() const
Tom Gundersen [Wed, 1 Apr 2015 11:46:59 +0000 (13:46 +0200)]
shared: set - make argument to set_remove() const

9 years agomount: fix up wording in the comment
Michal Sekletar [Wed, 1 Apr 2015 11:08:25 +0000 (13:08 +0200)]
mount: fix up wording in the comment

9 years agomount: don't run quotaon only for network filesystems
Lukas Nykryn [Mon, 30 Mar 2015 12:42:02 +0000 (14:42 +0200)]
mount: don't run quotaon only for network filesystems

If you have for example ext4 on iscsi devices it is possible to setup
qoutas there. Unfortunately, because such fstab entry contains _netdev,
systemd will not add dependency to quotaon.service.

9 years agohwdb: Update database of Bluetooth company identifiers
Marcel Holtmann [Tue, 31 Mar 2015 20:59:12 +0000 (22:59 +0200)]
hwdb: Update database of Bluetooth company identifiers

9 years agonspawn: fallback on bind mount when mknod fails
Alban Crequy [Tue, 31 Mar 2015 15:14:48 +0000 (17:14 +0200)]
nspawn: fallback on bind mount when mknod fails

Some systems abusively restrict mknod, even when the device node already
exists in /dev. This is unfortunate because it prevents systemd-nspawn
from creating the basic devices in /dev in the container.

This patch implements a workaround: when mknod fails, fallback on bind
mounts.

Additionally, /dev/console was created with a mknod with the same
major/minor as /dev/null before bind mounting a pts on it. This patch
removes the mknod and creates an empty regular file instead.

In order to test this patch, I used the following configuration, which I
think should replicate the system with the abusive restriction on mknod:

  # grep devices /proc/self/cgroup
  4:devices:/user.slice/restrict
  # cat /sys/fs/cgroup/devices/user.slice/restrict/devices.list
  c 1:9 r
  c 5:2 rw
  c 136:* rw
  # systemd-nspawn --register=false -D .

v2:
 - remove "bind", it is not needed since there is already MS_BIND
v3:
 - fix error management when calling touch()
 - fix lowercase in error message

9 years agonspawn: drop sd_booted() check
Lennart Poettering [Tue, 31 Mar 2015 13:34:01 +0000 (15:34 +0200)]
nspawn: drop sd_booted() check

We have no such check in any of the other tools, hence don't have one in
nspawn either.

(This should make things nicer for Rocket, among other things)

Note: removing this check does not mean that we support running nspawn
on non-systemd. We explicitly don't. It just means that we remove the
check for running it like that. You are still on your own if you do...

9 years agonspawn: change filesystem type from "bind" to NULL in mount() syscalls
Iago López Galeiras [Tue, 31 Mar 2015 09:50:29 +0000 (11:50 +0200)]
nspawn: change filesystem type from "bind" to NULL in mount() syscalls

Try to keep syscalls as minimal as possible.

9 years agoman: fix line breaks in code examples
Lennart Poettering [Tue, 31 Mar 2015 13:32:43 +0000 (15:32 +0200)]
man: fix line breaks in code examples

9 years agoARM: detect-virt: detect QEMU/KVM
Andrew Jones [Tue, 31 Mar 2015 09:08:13 +0000 (11:08 +0200)]
ARM: detect-virt: detect QEMU/KVM

QEMU/KVM guests do not have hypervisor nodes, but they do have
fw-cfg nodes (since qemu v2.3.0-rc0). fw-cfg nodes are documented,
see kernel doc Documentation/devicetree/bindings/arm/fw-cfg.txt,
and therefore we should be able to rely on it in this detection.

Unfortunately, we currently don't have enough information in the
DT, or elsewhere, to determine if we're using KVM acceleration
with QEMU or not, so we can only report 'qemu' at this time, even
if KVM is in use. This shouldn't really matter in practice though,
because if detect-virt is used interactively it will be clear to
the user whether or not KVM acceleration is present by the overall
speed of the guest. If used by a script, then the script's behavior
should not change whether it's 'qemu' or 'kvm'. QEMU emulated
guests and QEMU/KVM guests of the same type should behave
identically, only the speed at which they run should differ.

9 years agoARM: detect-virt: detect Xen
Andrew Jones [Tue, 31 Mar 2015 09:08:12 +0000 (11:08 +0200)]
ARM: detect-virt: detect Xen

9 years agodetect-virt: use /proc/device-tree
Andrew Jones [Tue, 31 Mar 2015 09:08:11 +0000 (11:08 +0200)]
detect-virt: use /proc/device-tree

Kernel doc Documentation/ABI/testing/sysfs-firmware-ofw says that
the /proc/device-tree symlink should be used, as opposed to
directly accessing /sys/firmware/devicetree/base. The former is
ABI, but not the later.

9 years agosystemd-bootchart: Repair Entropy Graph
Alexander Sverdlin [Sun, 29 Mar 2015 18:46:42 +0000 (20:46 +0200)]
systemd-bootchart: Repair Entropy Graph

Entropy Graph code doesn't handle the error condition if open() of /proc entry
fails. Moreover, the file is only opened once and only first sample will contain
the correct value because the return value of pread() is also not handled
properly and file is not re-opened. Fix both problems.

9 years agosystemd-bootchart: Prevent leaking file descriptors in open-fdopen combination
Alexander Sverdlin [Sun, 29 Mar 2015 18:44:04 +0000 (20:44 +0200)]
systemd-bootchart: Prevent leaking file descriptors in open-fdopen combination

Correctly handle the potential failure of fdopen() (because of OOM, for instance)
after potentially successful open(). Prevent leaking open fd in such case.

9 years agosystemd-bootchart: Prevent closing random file descriptors
Alexander Sverdlin [Sun, 29 Mar 2015 18:41:56 +0000 (20:41 +0200)]
systemd-bootchart: Prevent closing random file descriptors

If the kernel has no CONFIG_SCHED_DEBUG option set, systemd-bootchart produces
empty .svg file. The reason for this is very fragile file descriptor logic in
log_sample() and main() (/* do some cleanup, close fd's */ block). There are
many places where file descriptors are closed on failure (missing SCHED_DEBUG
provokes it), but there are several problems with it:
 - following iterations in the loop see that the descriptor is non zero and do
   not open the corresponding file again;
 - "some cleanup" code closes already closed files and the descriptors are reused
   already, in particular for resulting .svg file;
 - static "vmstat" and "schedstat" variables in log_sample() made the situation
   even worse.

These are the strace fragments:

[...]
close(7)                                = -1 EBADF (Bad file descriptor)
close(-1)                               = -1 EBADF (Bad file descriptor)
pread(7, 0xbea60a2c, 4095, 0)           = -1 EBADF (Bad file descriptor)
close(7)                                = -1 EBADF (Bad file descriptor)
close(-1)                               = -1 EBADF (Bad file descriptor)
pread(7, 0xbea60a2c, 4095, 0)           = -1 EBADF (Bad file descriptor)
close(7)                                = -1 EBADF (Bad file descriptor)
close(-1)                               = -1 EBADF (Bad file descriptor)
getdents64(4, /* 0 entries */, 32768)   = 0
clock_gettime(CLOCK_MONOTONIC, {24, 783843501}) = 0
nanosleep({0, 5221792}, NULL)           = 0
clock_gettime(CLOCK_MONOTONIC, {24, 789726835}) = 0
lseek(4, 0, SEEK_SET)                   = 0
pread(5, "nr_free_pages 52309\nnr_alloc_bat"..., 4095, 0) = 685
pread(6, "version 15\ntimestamp 4294939775\n"..., 4095, 0) = 86
getdents64(4, /* 99 entries */, 32768)  = 2680
pread(7, 0xbea60a2c, 4095, 0)           = -1 EBADF (Bad file descriptor)
close(7)                                = -1 EBADF (Bad file descriptor)
close(-1)                               = -1 EBADF (Bad file descriptor)
pread(8, 0xbea60a2c, 4095, 0)           = -1 EBADF (Bad file descriptor)
close(8)                                = -1 EBADF (Bad file descriptor)
close(-1)                               = -1 EBADF (Bad file descriptor)
pread(9, 0xbea60a2c, 4095, 0)           = -1 EBADF (Bad file descriptor)
close(9)                                = -1 EBADF (Bad file descriptor)
[...]

where it obviously tries to close same and reused decriptors many times, also
passing return code "-1" instead of descriptor...

[...]
close(7)                                = -1 EBADF (Bad file descriptor)
close(-1)                               = -1 EBADF (Bad file descriptor)
pipe2([7, 8], O_CLOEXEC)                = 0
clone(child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0xb6fd0068) = 192
close(8)                                = 0
fcntl64(7, F_SETFD, 0)                  = 0
fstat64(7, {st_mode=S_IFIFO|0600, st_size=0, ...}) = 0
mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0xb6fd2000
read(7, "[    0.074507] calling  vfp_init"..., 4096) = 4096
[...]
read(7, "s)\n[    6.228910] UBIFS: reserve"..., 4096) = 4096
read(7, "trary Executable File Formats Fi"..., 4096) = 1616
read(7, "", 4096)                       = 0
close(7)                                = 0
wait4(192, [{WIFEXITED(s) && WEXITSTATUS(s) == 0}], 0, NULL) = 192

9 years agotest: udev - add gdb support
Tom Gundersen [Wed, 25 Mar 2015 11:28:15 +0000 (12:28 +0100)]
test: udev - add gdb support

9 years agonetwork: don't pass NULL to udev_device_get_driver()
Tom Gundersen [Thu, 26 Mar 2015 20:53:38 +0000 (21:53 +0100)]
network: don't pass NULL to udev_device_get_driver()

9 years ago50-udev-default.rules: don't run anything but REMOVE_CMD on remove
Harald Hoyer [Fri, 27 Mar 2015 14:42:04 +0000 (15:42 +0100)]
50-udev-default.rules: don't run anything but REMOVE_CMD on remove

we don't want to run usb_id and input_id on ACTION=="remove"

9 years agocdrom_id: unroll and simplify data check loop
Harald Hoyer [Fri, 27 Mar 2015 12:47:32 +0000 (13:47 +0100)]
cdrom_id: unroll and simplify data check loop

also removes this warning:

src/udev/cdrom_id/cdrom_id.c: In function ‘cd_media_info.isra.13’:
src/udev/cdrom_id/cdrom_id.c:612:12: warning: assuming signed overflow
does not occur when assuming that (X + c) >= X is always true
[-Wstrict-overflow]
 static int cd_media_info(struct udev *udev, int fd)
            ^

9 years agofix gcc warnings about uninitialized variables
Harald Hoyer [Fri, 27 Mar 2015 11:02:49 +0000 (12:02 +0100)]
fix gcc warnings about uninitialized variables

like:

src/shared/install.c: In function ‘unit_file_lookup_state’:
src/shared/install.c:1861:16: warning: ‘r’ may be used uninitialized in
this function [-Wmaybe-uninitialized]
         return r < 0 ? r : state;
                ^
src/shared/install.c:1796:13: note: ‘r’ was declared here
         int r;
             ^

9 years agonetworkd-dhcp6: Do not handle prefix expiry
Patrik Flykt [Wed, 25 Mar 2015 11:22:43 +0000 (13:22 +0200)]
networkd-dhcp6: Do not handle prefix expiry

Expiring prefixes need not be handled anymore as the kernel has been
instructed not to create routes for DHCPv6 assigned addresses via the
IFA_F_NOPREFIXROUTE flag.

9 years agosystemd-networkd: Use IFA_F_NOPREFIXROUTE with IPv6 addresses
Patrik Flykt [Mon, 2 Feb 2015 11:13:17 +0000 (13:13 +0200)]
systemd-networkd: Use IFA_F_NOPREFIXROUTE with IPv6 addresses

The IFA_F_NOPREFIXROUTE flag prevents the kernel from creating new onlink
prefixes when a DHCPv6 IPv6 address with a prefix length is set from user
space. IPv6 routing will follow the onlink status from Router Advertisment
Prefix Information options or any manually set route, which is the correct
thing to do.

As this flag has a larger value than what fits into an unsigned char, update
the flag attribute to an uint32_t and set it with an IFA_FLAGS attribute
when writing netlink messages to the kernel.

9 years agosd-rtnl: Always enable IFA_FLAGS
Patrik Flykt [Wed, 25 Mar 2015 11:19:21 +0000 (13:19 +0200)]
sd-rtnl: Always enable IFA_FLAGS

IFA_FLAGS is a discrete value and has no preprocessor #define defined for
it. Fix this by always using the value.

9 years agoutil: rework word parsing and c unescaping code
Lennart Poettering [Mon, 23 Mar 2015 11:55:36 +0000 (18:55 +0700)]
util: rework word parsing and c unescaping code

When parsing words from input files, optionally automatically unescape
the passed strings, controllable via a new flags parameter.

Make use of this in tmpfiles, and port everything else over, too.

This improves parsing quite a bit, since we no longer have to process the
same string multiple times with different calls, where an earlier call
might corrupt the input for a later call.

9 years agoimportd: simplify expression
Lennart Poettering [Mon, 23 Mar 2015 11:54:58 +0000 (18:54 +0700)]
importd: simplify expression

9 years agotmpfiles: minor simplification
Lennart Poettering [Mon, 23 Mar 2015 11:54:31 +0000 (18:54 +0700)]
tmpfiles: minor simplification

9 years agoTODO: Nothing to be done for IPv6 DAD, handled by kernel
Patrik Flykt [Wed, 25 Mar 2015 14:00:36 +0000 (16:00 +0200)]
TODO: Nothing to be done for IPv6 DAD, handled by kernel

9 years agofstab-generator: don't accept missing root=, but accept root=none
Tobias Hunger [Tue, 24 Mar 2015 23:05:38 +0000 (00:05 +0100)]
fstab-generator: don't accept missing root=, but accept root=none

And other non-device entries (like fstab does).

Mount whatever the user asked to be mounted on / on the kernel
command line. Do less sanity check and do *not* bail out
when the mount device looks strange or does not exist.

This basically makes the changes for deviceless filesystems
from yesterday unnecessary and is in line with what we do for
filesystems set up in fstab.

Remove some code that is now dead (reverting fb02a2775a65 and
b0438462).

[tomegun:
  - change patch title/description a bit.
  - don't touch the /usr logic, that would be a separate change and
    we don't currently have a convincing use-case for that.
  - don't bail out on /sys ro. This only makes sense in containers,
    where we would not be doing this anyway. If there is a use-case
    we could consider that as a separate patch.]

9 years agotmpfiles: avoid out of bounds read
Thomas Hindoe Paaboel Andersen [Tue, 24 Mar 2015 22:37:48 +0000 (23:37 +0100)]
tmpfiles: avoid out of bounds read

Otherwise this will go wrong for 'v'.

9 years agorules: storage - support MemoryStick (non-Pro) cards
Mantas Mikulėnas [Tue, 17 Mar 2015 16:00:24 +0000 (18:00 +0200)]
rules: storage - support MemoryStick (non-Pro) cards

These are handled by a different driver than MemoryStick Pro.

9 years agorules: storage - whitelist partitioned MS & MMC devices
Kay Sievers [Tue, 24 Mar 2015 22:28:25 +0000 (23:28 +0100)]
rules: storage - whitelist partitioned MS & MMC devices

On Mon, Mar 23, 2015 at 8:55 AM, Mantas Mikulėnas <grawity@gmail.com> wrote:
> On Tue, Mar 17, 2015 at 11:50 PM, Kay Sievers <kay@vrfy.org> wrote:
>> On Tue, Mar 17, 2015 at 5:00 PM, Mantas Mikulėnas <grawity@gmail.com>
>> wrote:
>> > Accidentally dropped in 1aff20687f4868575.
>> > ---
>> >  rules/60-persistent-storage.rules | 2 +-
>> >  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> > +KERNEL!="loop*|mmcblk[0-9]*|mspblk[0-9]*|nvme*|sd*|sr*|vd*",
>> > GOTO="persistent_storage_end"
>>
>> We can't do that, we need to ignore the mmc*rpmb devices:
>>
>> http://cgit.freedesktop.org/systemd/systemd/commit/?id=b87b01cf83947f467f3c46d9831cd67955fc46b9
>>
>> Maybe "mmcblk*[0-9]" will work?
>
> Yeah, that would probably work (the names are like mmcblk0p1 etc.)

9 years agoTODO
Tom Gundersen [Tue, 24 Mar 2015 22:05:43 +0000 (23:05 +0100)]
TODO

9 years agotimedatectl: remove unused variables
Thomas Hindoe Paaboel Andersen [Tue, 24 Mar 2015 21:55:31 +0000 (22:55 +0100)]
timedatectl: remove unused variables

9 years agoman: Second example systemd-socket-proxyd does not need a Unix socket listener.
David Strauss [Tue, 24 Mar 2015 19:25:56 +0000 (12:25 -0700)]
man: Second example systemd-socket-proxyd does not need a Unix socket listener.

9 years agorules: storage - don't apply rules to remove events
Tom Gundersen [Tue, 24 Mar 2015 14:15:45 +0000 (15:15 +0100)]
rules: storage - don't apply rules to remove events

This line was accidentally lost in 52346b5f5424.

9 years agotimedate: remove daylight saving time handling and tzfile parser
Kay Sievers [Tue, 24 Mar 2015 12:52:04 +0000 (13:52 +0100)]
timedate: remove daylight saving time handling and tzfile parser

We planned to support (the conceptually broken) daylight saving
time/local time features in the kernel, SCSI, networking, FAT
filesystem, but it turned out to be a race we cannot win and do
not want to get involved. Systemd should not fiddle with daylight
saving time or parse timezone information itself.

Leave everything to glibc or tools like date(1) and do not make any
promises or raise expectations that systemd should handle anything
like this.

9 years agotimedatectl: fix when queried system has differn't timezone
Shawn Landden [Mon, 23 Mar 2015 11:44:57 +0000 (04:44 -0700)]
timedatectl: fix when queried system has differn't timezone

Also allow getting time from time(2) when BUS_TRANSPORT_MACHINE.

v2: check for error

9 years agopo: update Russian translation
Sergey Ptashnick [Sun, 22 Mar 2015 19:00:13 +0000 (22:00 +0300)]
po: update Russian translation

Add strings for importd, following 587fec427c.

9 years agobootchart: more useful error message for common error
Shawn Landden [Mon, 23 Mar 2015 10:51:35 +0000 (03:51 -0700)]
bootchart: more useful error message for common error

Reported-by: tfirg_ on IRC
9 years agobootchart: remove duplicated code, prevent creating empty files
Shawn Landden [Mon, 23 Mar 2015 10:48:41 +0000 (03:48 -0700)]
bootchart: remove duplicated code, prevent creating empty files

In Debian and rawhide Fedora, which have CONFIG_SCHEDSTATS=n,
bootchart creates empty files in /run/log before printing an error.
Stop doing that.

Moreover this duplicated part of the code doesn't even have error checking
so there is no error avoided by doing this early.

Reported-by: tfirg_ on IRC
9 years agofstab-generator: Do not check deviceless filesystems
Tobias Hunger [Mon, 23 Mar 2015 23:53:00 +0000 (00:53 +0100)]
fstab-generator: Do not check deviceless filesystems

There is no need to check those.

9 years agofstab-generator: Support root on tmpfs (or other deviceless FS)
Tobias Hunger [Mon, 23 Mar 2015 23:52:59 +0000 (00:52 +0100)]
fstab-generator: Support root on tmpfs (or other deviceless FS)

This allows for stateless systems.

9 years agosd-rtnl: handle empty multi-part message from the kernel
Alin Rauta [Wed, 18 Mar 2015 12:06:19 +0000 (05:06 -0700)]
sd-rtnl: handle empty multi-part message from the kernel

We strips out NLMSG_DONE piece from a multi-part message adding into the
receive queue only the messages containing actual data.

If we send a request to the kernel for getting the forwarding database table (just an example),
the response will be a multi-part message like below:
1. FDB entry 1;
2. FDB entry 2;
3. NLMSG_DONE;

We strip out "3. NLMSG_DONE;" part and places into the receive queue a pointer to
"1. FDB entry 1; 2. FDB entry 2".

But if the FDB table is empty, the respose from the kernel will look like below:
1. NLMSG_DONE;

We strip out "1. NLMSG_DONE;" part and since there is no actual data got, it continues
waiting until reaching timeout.

Therefore, a call to "sd_rtnl_call" to send and wait for a response from kernel will exit
with timeout which is interpreted as error in communication.

This patch puts the NLMSG_DONE message on the receive queue if it ends an empty multi-part
message. This situation is detected in sd_rtnl_call() and in the callback code and NULL is
returned to the caller instead.

[tomegun:
  - added/reworded commit message
  - extend the same support to sd_rtnl_call_async()
  - drop debug logging from library, we only do this if something is really wrong, but an
    empty multi-part message is perfectly normal
  - modernize the code we touch whilst we are at it]

9 years agoman: add journal-remote.conf(5)
Chris Morgan [Sun, 22 Mar 2015 00:47:46 +0000 (20:47 -0400)]
man: add journal-remote.conf(5)

9 years agotimedated: fix enable/disable reversal
Zbigniew Jędrzejewski-Szmek [Sat, 21 Mar 2015 23:21:17 +0000 (19:21 -0400)]
timedated: fix enable/disable reversal

Bug introduced in 984f1b1d1b. The state was flipped later,
but the enable/disable routine made use of the state to decide
what to do.

context_enable_ntp() and context_start_ntp() now get the desired
state directly, so the Context parameter can be removed.

9 years agocore: make SELinux enable/disable check symmetric
Zbigniew Jędrzejewski-Szmek [Sat, 21 Mar 2015 22:50:10 +0000 (18:50 -0400)]
core: make SELinux enable/disable check symmetric

We'd use the generic check for disable, and a unit-file-specific one for enable.
Use the more specific one both ways.

systemd[1]: SELinux access check scon=system_u:system_r:systemd_timedated_t:s0 tcon=system_u:system_r:init_t:s0 tclass=system perm=disable path=(null) cmdline=/usr/lib/systemd/systemd-timedated: -13
systemd[1]: SELinux access check scon=system_u:system_r:systemd_timedated_t:s0 tcon=system_u:object_r:systemd_unit_file_t:s0 tclass=service perm=enable path=/usr/lib/systemd/system/systemd-timesyncd.service cmdline=/usr/lib/systemd/systemd-timedated: -13

https://bugzilla.redhat.com/show_bug.cgi?id=1014315

9 years agotimedated: flip internal status after executing operation
Zbigniew Jędrzejewski-Szmek [Sat, 21 Mar 2015 21:40:20 +0000 (17:40 -0400)]
timedated: flip internal status after executing operation

timedated would set the internal status before calling out to systemd to do
the actual change. When the operation was refused because of a SELinux denial,
the state kept in timedated would get out of sync, and the second call from
timedatectl would appear to succeed.

https://bugzilla.redhat.com/show_bug.cgi?id=1014315

9 years agomissing.h: add more btrfs types and defines
Michael Olbrich [Wed, 18 Mar 2015 13:04:55 +0000 (14:04 +0100)]
missing.h: add more btrfs types and defines

9 years agofstab-generator: ignore invalid swap priority
Zbigniew Jędrzejewski-Szmek [Sat, 21 Mar 2015 15:31:16 +0000 (11:31 -0400)]
fstab-generator: ignore invalid swap priority

A failed priority is not something worth stopping boot over. Most people
have only one swap device, in which case priority is irrelevant, and even
if there is more than one swap device, they are all usable, and ignoring the
priority field should only result in some loss of performance.

The kernel will report the priority as -1 if not set, so it's easy for
people to make this mistake.

https://bugzilla.redhat.com/show_bug.cgi?id=1204336

9 years agonetworkd: use LOG_WARN not LOG_ERR for non-fatal errors
Zbigniew Jędrzejewski-Szmek [Mon, 16 Mar 2015 15:35:12 +0000 (11:35 -0400)]
networkd: use LOG_WARN not LOG_ERR for non-fatal errors

9 years agoshared/mkdir: do not set errno in addition to return value
Zbigniew Jędrzejewski-Szmek [Mon, 16 Mar 2015 15:34:39 +0000 (11:34 -0400)]
shared/mkdir: do not set errno in addition to return value

All callers look at the return value anyway.

9 years agorules: avoid 'device/' accesses
David Herrmann [Fri, 20 Mar 2015 11:28:12 +0000 (12:28 +0100)]
rules: avoid 'device/' accesses

We should never access parents, as the sysfs hierarchy is in no way
stable. Use KERNELS== etc. to match on a parent, then access it via
$attr{} (which accesses the matching device, not the current device).

9 years agorules: fix input-name for keyboard rules
David Herrmann [Fri, 20 Mar 2015 09:44:21 +0000 (10:44 +0100)]
rules: fix input-name for keyboard rules

We match on the evdev node, but only the parent has a "name" attribute.
Use $attr{device/name} to access it.

This is borked since 2013, I wonder how that ever worked? Maybe this will
suddenly fix all the DMI-based key detections.

Thanks to Peter Hutterer for catching this!

9 years agoman: add h/H to tmpfiles.d(5)
Goffredo Baroncelli [Mon, 16 Mar 2015 19:33:51 +0000 (20:33 +0100)]
man: add h/H to tmpfiles.d(5)

9 years agoAllow systemd-tmpfiles to set the file/directory attributes
Goffredo Baroncelli [Mon, 16 Mar 2015 19:33:50 +0000 (20:33 +0100)]
Allow systemd-tmpfiles to set the file/directory attributes

Allow systemd-tmpfiles to set the file/directory attributes, like
chattr(1) does. Two more commands are added: 'H' and 'h' to set the
attributes, recursively and not.

9 years agoAdd change_attr_fd()
Goffredo Baroncelli [Mon, 16 Mar 2015 19:33:49 +0000 (20:33 +0100)]
Add change_attr_fd()

Add change_attr_fd() function to modify the file/directory attribute.

9 years agopo: updated Swedish translation
Sebastian Rasmussen [Fri, 20 Mar 2015 01:55:25 +0000 (21:55 -0400)]
po: updated Swedish translation

https://bugs.freedesktop.org/show_bug.cgi?id=89598

9 years agohwdb: ship ids-update.pl & sdio.ids in the release tarballs
Dimitri John Ledkov [Mon, 16 Mar 2015 22:54:07 +0000 (22:54 +0000)]
hwdb: ship ids-update.pl & sdio.ids in the release tarballs

This makes it easier to apply stable branch patches on top of the
release tarball.

9 years agohwdb: add HP ProBook 450G0
Zbigniew Jędrzejewski-Szmek [Thu, 19 Mar 2015 23:48:57 +0000 (19:48 -0400)]
hwdb: add HP ProBook 450G0

https://bugs.freedesktop.org/show_bug.cgi?id=89684

9 years agofsck: use %zu for size_t
Thomas Hindoe Paaboel Andersen [Wed, 18 Mar 2015 21:36:16 +0000 (22:36 +0100)]
fsck: use %zu for size_t

cur and max were changed to size_t in ac6e2f0dfc2e800f01c79ecea2b811914110a7b9
update the format accordingly.

9 years agoresolved: Do not add .busname dependencies, when compiling without kdbus.
Dimitri John Ledkov [Tue, 17 Mar 2015 15:37:07 +0000 (16:37 +0100)]
resolved: Do not add .busname dependencies, when compiling without kdbus.

9 years agostrv: return NULL from strv_free()
David Herrmann [Tue, 17 Mar 2015 11:20:31 +0000 (12:20 +0100)]
strv: return NULL from strv_free()

We always return NULL/invalid-object from destructors, fix strv_free() to
do the same.

9 years agobootctl: drop redundant parse_argv()
David Herrmann [Mon, 16 Mar 2015 20:07:38 +0000 (21:07 +0100)]
bootctl: drop redundant parse_argv()

We already call parse_argv() from main(), don't call it here again.

9 years agolibudev: private - introduce udev_device_new_from_synthetic_event()
Tom Gundersen [Sun, 15 Mar 2015 16:10:17 +0000 (17:10 +0100)]
libudev: private - introduce udev_device_new_from_synthetic_event()

This allows set_action(), read_uevent_file() and read_db() to be made internal to libudev.

9 years agoconsole-getty.service: don't start when /dev/console is missing
Jan Pazdziora [Fri, 13 Mar 2015 11:57:18 +0000 (12:57 +0100)]
console-getty.service: don't start when /dev/console is missing

Create minimal image which runs systemd

   FROM rhel7.1
   RUN yum install -y /usr/bin/ps
   ENV container docker
   CMD [ "/usr/sbin/init" ]

When you run the container without -t, the process

   /sbin/agetty --noclear --keep-baud console 115200 38400 9600

is not happy and checking the journal in the container, there is a stream of

Mar 13 04:50:15 11bf07f59fff agetty[66]: /dev/console: No such file or directory
Mar 13 04:50:25 11bf07f59fff systemd[1]: console-getty.service holdoff time over, scheduling restart.
Mar 13 04:50:25 11bf07f59fff systemd[1]: Stopping Console Getty...
Mar 13 04:50:25 11bf07f59fff systemd[1]: Starting Console Getty...
Mar 13 04:50:25 11bf07f59fff systemd[1]: Started Console Getty.
Mar 13 04:50:25 11bf07f59fff agetty[67]: /dev/console: No such file or directory
Mar 13 04:50:35 11bf07f59fff systemd[1]: console-getty.service holdoff time over, scheduling restart.
Mar 13 04:50:35 11bf07f59fff systemd[1]: Stopping Console Getty...
Mar 13 04:50:35 11bf07f59fff systemd[1]: Starting Console Getty...
Mar 13 04:50:35 11bf07f59fff systemd[1]: Started Console Getty.
Mar 13 04:50:35 11bf07f59fff agetty[74]: /dev/console: No such file or directory
Mar 13 04:50:45 11bf07f59fff systemd[1]: console-getty.service holdoff time over, scheduling restart.
Mar 13 04:50:45 11bf07f59fff systemd[1]: Stopping Console Getty...
Mar 13 04:50:45 11bf07f59fff systemd[1]: Starting Console Getty...

9 years agocore/namespace: fix path sorting
Michal Schmidt [Mon, 16 Mar 2015 21:04:21 +0000 (22:04 +0100)]
core/namespace: fix path sorting

The comparison function we use for qsorting paths is overly indifferent.
Consider these 3 paths for sorting:
 /foo
 /bar
 /foo/foo
qsort() may compare:
 "/foo" with "/bar" => 0, indifference
 "/bar" with "/foo/foo" => 0, indifference
and assume transitively that "/foo" and "/foo/foo" are also indifferent.

But this is wrong, we want "/foo" sorted before "/foo/foo".
The comparison function must be transitive.

Use path_compare(), which behaves properly.

Fixes: https://bugzilla.redhat.com/show_bug.cgi?id=1184016
9 years agoshared: add path_compare(), an ordering path comparison
Michal Schmidt [Mon, 16 Mar 2015 20:58:35 +0000 (21:58 +0100)]
shared: add path_compare(), an ordering path comparison

... and make path_equal() a simple wrapper around it.

9 years agorules: keyboard - prefix "atkbd" match strings like we prefix the "name" strings
Kay Sievers [Mon, 16 Mar 2015 17:43:14 +0000 (18:43 +0100)]
rules: keyboard - prefix "atkbd" match strings like we prefix the "name" strings

9 years agorules: keyboard - only search the "input" subsystem for the modalias
Kay Sievers [Mon, 16 Mar 2015 17:33:45 +0000 (18:33 +0100)]
rules: keyboard - only search the "input" subsystem for the modalias

9 years agofsckd: check if plymouth is running before attempting connection
Didier Roche [Tue, 10 Mar 2015 09:05:19 +0000 (10:05 +0100)]
fsckd: check if plymouth is running before attempting connection

9 years agofsckd: Don't use strjoina on gettext() call
Didier Roche [Tue, 10 Mar 2015 07:58:23 +0000 (08:58 +0100)]
fsckd: Don't use strjoina on gettext() call

9 years agoshared - reinstate plymouth_running()
Tom Gundersen [Mon, 16 Mar 2015 17:29:26 +0000 (18:29 +0100)]
shared - reinstate plymouth_running()

This will be used by fsckd. This reverts part of
031886edfc6e96ab778c241035a8d00fb0de99d3.

9 years agocore: Remove explicit Plymouth integration
Jasper St. Pierre [Mon, 16 Mar 2015 16:34:59 +0000 (09:34 -0700)]
core: Remove explicit Plymouth integration

Even if plymouth is running, it might have not displayed the splash yet,
so we'll see a few lines on fbcon when we should have otherwise had
nothing.

Plymouth integration was added to systemd in commit
6faa11140bf776cdaeb8d22d01816e6e48296971. That same day, Plymouth got
systemd integration [0]. As such, the Plymouth integration has always
been obsolete, and was probably only for older Plymouth's. But I can't
imagine anybody running a Plymouth from 2011 with a systemd from 2015.

Remove the Plymouth/systemd integration, and let Plymouth's code tell
systemd to print the details.

[0] http://cgit.freedesktop.org/plymouth/commit/?id=537c16422cd49f1beeaab1ad39846a00018faec1

Signed-off-by: Jasper St. Pierre <jstpierre@mecheye.net>
Cc: Daniel Drake <dsd@endlessm.com>
Cc: Ray Strode <rstrode@redhat.com>
9 years agoRevert "hwdb: merge atkbd into platform matches"
David Herrmann [Mon, 16 Mar 2015 14:51:25 +0000 (15:51 +0100)]
Revert "hwdb: merge atkbd into platform matches"

This reverts commit ba76ee29bc02879fb42c048132af8889b00220d5. As it turns
out, we need to match on driver=atkbd to not load the fixups on any
plugged USB devices.

That is, whenever you use "name:<name>:dmi:<dmi>" style matches, you
better provide a name or you're screwing things up.

9 years agocore: don't change removed devices to state "tentative"
Martin Pitt [Fri, 13 Mar 2015 07:35:59 +0000 (08:35 +0100)]
core: don't change removed devices to state "tentative"

Commit 628c89c introduced the "tentative" device state, which caused
devices to go from "plugged" to "tentative" on a remove uevent. This
breaks the cleanup of stale mounts (see commit 3b48ce4), as that only
applies to "dead" devices.

The "tentative" state only really makes sense on adding a device when
we don't know where it was coming from (i. e. not from udev). But when
we get a device removal from udev we definitively know that it's gone,
so change the device state back to "dead" as before 628c89c.

9 years agorules: keyboard - remove "platform" from comments
Kay Sievers [Mon, 16 Mar 2015 12:43:50 +0000 (13:43 +0100)]
rules: keyboard - remove "platform" from comments

9 years agohwdb: merge atkbd into platform matches
David Herrmann [Mon, 16 Mar 2015 11:19:09 +0000 (12:19 +0100)]
hwdb: merge atkbd into platform matches

Currently, we always run
  hwdb 'keyboard:name:$attr{name}:$attr{[dmi/id]modalias}'
as last step to match keyboards. Therefore, if nothing else matched so
far, we still try the device-name+dmi combination.

However, we have a special atkbd rule which is only run for atkbd as:
  hwdb 'keyboard:$attr{[dmi/id]modalias}'

This is redundant, as we already pass the same information to hwdb in the
last fallback step.

This patch converts the hwdb "keyboard:dmi:*" matches to
"keyboard:name:*:dmi:*" matches and drops the redundant rule.

9 years agorules: drop redundant match
David Herrmann [Mon, 16 Mar 2015 11:05:36 +0000 (12:05 +0100)]
rules: drop redundant match

The 60-keyboard rules are already guared by KERNEL!="event*" bail-outs,
therefore, KERNELS="input*" is always true. Drop it!

9 years agohwdb: support bluetooth keyboard fixups
David Herrmann [Mon, 16 Mar 2015 11:04:38 +0000 (12:04 +0100)]
hwdb: support bluetooth keyboard fixups

Drop the restriction not to match on bluetooth devices. They are supported
just fine!

9 years agohwdb: convert to generic input-modalias matches
David Herrmann [Mon, 16 Mar 2015 11:00:46 +0000 (12:00 +0100)]
hwdb: convert to generic input-modalias matches

There is no reason to match on usb-modaliases, if we can use the
input-modalias to achieve the same. This commit changes the
keyboard-lookups to not be restricted to USB, but pass all modaliases to
the hwdb. Furthermore, we convert all usb:* matches to input:* matches,
thus getting rid of any ambiguity if multiple usb devices are chained (or
a bluetooth device / etc. is on top).

Note that legacy keyboard:usb:* matches are still supported, but
deprecated. If possible, please use keyboard:input:* matches instead.

This is a required step to make other input devices work with
60-keyboard.hwdb. Other bus-types are often chained on usb and we want to
avoid any ambiguity here if we incorrectly match on a USB hub.