Kay Sievers [Wed, 3 Jun 2009 23:44:04 +0000 (01:44 +0200)]
udevd: convert to event worker processes
Event processes now get re-used after they handled an event. This reduces
pressure on the CPU significantly because cloned event processes no longer
cause page faults in the main daemon. After the events have settled, the
no longer needed worker processes get killed.
Benjamin Gilbert [Fri, 22 May 2009 18:54:46 +0000 (20:54 +0200)]
test: check string substitutions in OWNER and GROUP
String substitutions in OWNER and GROUP keys were broken in udev 137-142.
Explicitly test for this, since such breakage will not manifest in typical
rulesets.
Kay Sievers [Fri, 22 May 2009 14:44:53 +0000 (16:44 +0200)]
rules: add SUBSYSTEM match to scsi rules
On Fri, May 22, 2009 at 16:15, Alan Jenkins <alan-jenkins@tuffmail.co.uk> wrote:
> I've been looking at what is responsible for all the path lookup activity in
> coldplug. On my debian stable system, it looks like every device gets its
> parent looked up in sysfs. I think this is due to SUBSYSTEMS matches.
>
> I see the udev default rules are different, but it looks like they still
> test for SUBSYSTEMS on every single device. Should we add SUBSYSTEM="scsi_generic"
> to these three rules?
Alan Jenkins [Thu, 21 May 2009 20:22:37 +0000 (22:22 +0200)]
udevd: implement a more efficient queue file format
Directory lookups show up in profiling. The queue files are responsible
for a large proportion of file-related system calls in udev coldplug.
Instead of creating a file for each event, append their details to a
log file. The file is periodically rebuilt (garbage-collected) to
prevent it from growing indefinitely.
This single queue file replaces both the queue directory and the
uevent_seqnum file. On desktop systems the file tends not to grow
beyond one page. So it should also save a small amount of memory in
tmpfs.
Tests on a running EeePC indicate average savings of 5% *udevd* cpu time
as measured by oprofile. __link_path_walk is reduced from 1.5% to
1.3%. It is not completely clear where the rest of the gains come from.
In tests running ~400 events, the queue file is rebuilt about 5 times.
Signed-off-by: Alan Jenkins <alan-jenkins@tuffmail.co.uk>
The introduction of the --resolve-names=early/never code introduced a
bug to the OWNER/GROUP lookup. Previously if the name had contained $,
lookup would have been performed later; after the patch, the key ended
up being ignored!
Kay Sievers [Thu, 7 May 2009 11:24:40 +0000 (13:24 +0200)]
add "static" back to the inline functions
<Keybuk> kay: udev git head ftbfs
<Keybuk> udev-watch.o: In function `udev_selinux_init':
<Keybuk> /../udev/udev.h:130: multiple definition of `udev_selinux_init'
Kay Sievers [Mon, 4 May 2009 02:52:31 +0000 (04:52 +0200)]
handle devtmpfs nodes
UDev follows the kernel given name, and re-uses the kernel created
device node. If the kernel and spcecified udev rules disagree, the
udev specified node node is created and the kernel-created on is
deleted.
Lubomir Rintel [Sun, 3 May 2009 13:37:30 +0000 (15:37 +0200)]
rule-generator: net - whitelist NICs that violate MAC local scheme
According to list of assigned ethernet codes [1] referred to by
IANA [2] certain global addresses do not follow the assignement
scheme and use numbers reserved for local use. Several such adapters
are quite widely used, generate rules for them.
Kay Sievers [Sun, 26 Apr 2009 18:53:30 +0000 (20:53 +0200)]
rule_generator: net - fix usb comment generation
On Sat, Apr 4, 2009 at 22:17, Omair Eshkenazi <stimpson@phys.huji.ac.il> wrote:
> I noticed that in (70-)persistent-net.rules, the comments for USB devices
> are missing the device/vendor id's. Example:
> # USB device 0x:0x (rt73usb)
Andre Przywara [Thu, 23 Apr 2009 22:29:53 +0000 (00:29 +0200)]
rules: create /dev/cpu/<n>/cpuid world readable
I don't see any security implications, to be actually useful,
/dev/cpu/<n>/cpuid should be world readable. The cpuid instruction
can be called from userspace anyway, so there is nothing to hide.
The device does not support any write operation, so 0444 should
suffice.
Signed-off-by: Andre Przywara <andre.przywara@amd.com>
Kay Sievers [Thu, 23 Apr 2009 11:23:45 +0000 (13:23 +0200)]
libudev: queue - fix get_seqnum_is_finished()
On Thu, Apr 23, 2009 at 11:39, Alan Jenkins <alan-jenkins@tuffmail.co.uk> wrote:
> I'm puzzled by this function:
>
> /* if we have not seen this seqnum, check if it is/was already queued */
> if (seqnum < udev_queue->last_seen_udev_seqnum) {
> udev_queue_get_udev_seqnum(udev_queue);
> if (seqnum < udev_queue->last_seen_udev_seqnum)
>
> Shouldn't the test be (seqnum > udev_queue->last_seen_udev_seqnum) ?
John Wright [Wed, 22 Apr 2009 12:53:35 +0000 (14:53 +0200)]
edd_id: add cciss devices
Since cciss devices can be BIOS boot devices just as well as sd* and
hd*, the edd_id program should be run on them so that the later
/dev/disk/by-id/edd-* rules will work.
Kay Sievers [Wed, 22 Apr 2009 01:50:11 +0000 (03:50 +0200)]
libudev: monitor - add client socket filter for subsystem value
Messages send back by the udev daemon to the netlink socket are
multiplexed by the kernel and delivered to multiple clients. The
clients can upload a socket filter to let the kernel drop messages
not belonging to a certain subsystem. This prevent needless wakeups
and message processing for users who are only interested in a
subset of available events.
Recent kernels allow untrusted users to listen to the netlink
messages.
The messages send by the udev daemon are versioned, to prevent any
custom software reading them without libudev. The message wire format
may change with any udev version update.