Controls the behavior of the running daemon. Currently only stopping and starting
of the execution queue is supported.
Signed-off-by: Kay Sievers <kay.sievers@suse.de>
On Tue, 2005-03-15 at 09:25 +0100, Hannes Reinecke wrote:
> The current implementation of the firmware class breaks a fundamental
> assumption in udevd: that the physical device can be initialised fully
> prior to executing the next event for that device.
Thanks to Hannes for the patch.
If the system reaches a defined limit of processes in running state, udevd
starts to count its own processes in running state from its session (all
forked hotplug child processes, subprocesses and callouts) and throttles
further process forking if the limit is reached.
This should help setups with hundreds of events emitted hotplug events
in parallel with hundreds of processes in "R" state. which makes the machine
unresponsible.
I placed a 100% cpu time consuming program in /etc/hotplug.d/ which runs for 5
seconds. With this patch I can load "scsi_debug add_host=100" without any major
problem. Without the patch the box is completly unresponsible for many minutes.
Define an inititialization phase, where we delay the events only
for a short time to eventually recollect the event sequence instead
of delaying the very first events for 10 seconds, which breaks the
firmware loader.
Currently udevd delays only events for the same DEVPATH.
Example of an "add" event sequence:
/block/sda
/block/sda/sda1
With this change, we make sure, that the udev process handling
/block/sda has finished its work (waited for all attributes,
created the node) before we fork the udev event for /block/sda/sda1.
This way the event for sda1 can be sure, that the node for the
main device is already created (may be useful for disk labels).
It will not affect any parallel device handling, only the sequence
of the devices directory chain is serialized. The 10.000 disks
plugged in will still run as parallel events. :)
The main motivation to do this is the program execution of the
dev.d/ and hotplug.d/ directory. If we don't wait for the parent
event to exit, we can't be sure that the executed scripts are
run in the right order.
On Thu, Dec 09, 2004 at 09:18:28AM +0100, Kay Sievers wrote:
> On Wed, 2004-12-08 at 19:07 -0800, David Brownell wrote:
> > Could that argument apply to the underlying hardware, too?
> We now make sure that the sequence of events for a device
> is serialized for every device chain and the class/block
> devices which have a "device" link to a physical device are
> handled after the physical device is fully populated and
> notified to userspace. It will only work this way on kernels
> later than 2.6.10-rc1 cause it depends on the PHYSDEVPATH
> value in the hotplug environment.
Make _all_ hotplug variables available to the forked udev,
the udev callouts and the udev dev.d/ scripts. We put the
whole environment into a buffer and send it over the udevd
socket. udevd recreates *envp[] and passes it to the exec().
Change to the same timeout loop we use in the rest of the code. Change
some comments and names to be more descriptive.
I'm mostly finished with the overall cleanup. I will post a new patch
for the udevd-nofork experiment, which will be much smaller now.
here we change udevd to pass the SEQNUM from the hotplug environment
to udev and the dev.d/ scripts. We need this for HAL to match the
hotplug event with the dev.d/ events.
It also changes the type from int to long to match the kernel.
here is a patch on top of your nice improvements.
I fixed the whitespace and it hopefully fixes the stupid timestamp bug in
udevd. Some stupid OS sets the hwclock to localtime and linux changes it
to UTC while starting. If any events are pending they may be delayed by
the users time distance from UTC :) So we use the uptime seconds now.
Here we truncate our input strings from the environment to our
defined limit. It's a bit theroretical but better check for it.
It cleans up some magic length definitions and removes the code
duplication in udev, udevtest and udevsend.
udevd needs to be killed after installation, cause the message size
is changed with this patch.
Should we do this with the 'make install', like we do with the '.udevdb'?
As Chris Friesen <chris_friesen@sympatico.ca> suggested, here we switch
the unix domains socket path to abstract namespace and get rid of the
socket file in the filesystem.
Hey, this was new to me today. So here a few words:
Linux supports a abstract namespace for sockets. We don't need a
physical file on the filesystem but only a unique string magically
starting with the '\0' character.
strace with real file:
connect(3, {sa_family=AF_UNIX, path="/udev/.udevd.sock"}, 110)
strace with abstract namespace:
connect(3, {sa_family=AF_UNIX, path=@udevd}, 110)
> Here is a small cleanup and better Makefile integration.
> udevd and udevsender are now installed. Just switch HOTPLUG_EXEC from ROOT
> to SENDER before install and udevsend will be called.
>
> We may add the location of the socket and lock file to the config,
> if this is needed.
Same patch with a fix for the stack size setting.
On Thu, Jan 29, 2004 at 04:55:11PM +0100, Kay Sievers wrote:
> On Thu, Jan 29, 2004 at 02:56:25AM +0100, Kay Sievers wrote:
> > On Wed, Jan 28, 2004 at 10:47:36PM +0100, Kay Sievers wrote:
> > > Oh, couldn't resist to try threads.
> > > It's a multithreaded udevd that communicates through a localhost socket.
> > > The message includes a magic with the udev version, so we don't accept
> > > older udevsend's.
> > >
> > > No need for locking, cause we can't bind two sockets on the same address.
> > > The daemon tries to connect and if it fails it starts the daemon.
> > >
> > > We create a thread for every incoming connection, handle over the socket,
> > > sort the messages in the global message queue and exit the thread.
> > > Huh, that was easy with threads :)
> > >
> > > With the addition of a message we wakeup the queue manager thread and
> > > handle timeouts or move the message to the global exec list. This wakes
> > > up the exec list manager who looks if a process is already running for this
> > > device path.
> > > If yes, the exec is delayed otherwise we create a thread that execs udev.
> > > n the background. With the return of udev we free the message and wakeup
> > > the exec list manager to look if something is pending.
> > >
> > > It is just a quick shot, cause I couldn't solve the problems with fork an
> > > scheduling and I wanted to see if I'm to stupid :)
> > > But if anybody with a better idea or more experience with I/O scheduling
> > > we may go another way. The remaining problem is that klibc doesn't support
> > > threads.
> > >
> > > By now, we don't exec anything, it's just a sleep 3 for every exec,
> > > but you can see the queue management by watching syslog and do:
> > >
> > > DEVPATH=/abc ACTION=add SEQNUM=0 ./udevsend /abc
>
> Next version, switched to unix domain sockets.
Next cleaned up version. Hey, nobody wants to try it :)
Works for me, It's funny if I connect/disconnect my 4in1-usb-flash-reader
every two seconds. The 2.6 usb rocks! I can connect/diconnect a hub with 3
devices plugged in every second and don't run into any problem but a _very_
big udevd queue.
Here is the next round. We have three queues now. All incoming messages
are queued in msg_list and if nothing is missing we move it to the
running_list and exec in the background.
If the exec comes back, it removes the message from the running_list and
frees the message.
Before we exec, we check the running_list if there is a udev running on
the same device path. If yes, we move the message to the delay_list. If
the former exec comes back, we move the message to the running_list and
exec it.
The very first event is delayed now to catch possible earlier sequences,
every following event is executed without delay if no sequence is missing.
The daemon doesn't exit by itself any longer, cause we don't want to
delay every first exec.
I've put a $(PWD) for now in the Makefile for testing this beast. Only
the local binaries are executed, not the /sbin/udev. We can change it
if we are ready for real testing.
And SIGKILL can't be cought, so I removed it from the handler :)
06:58:36 sig_handler: caught signal 15
06:58:36 main: using ipc queue 0x2d548
06:58:37 message is still in the ipc queue, starting daemon...
06:58:37 work: received sequence 3, expected sequence 0
06:58:37 msg_dump_queue: sequence 3 in queue
06:58:37 set_timeout: set timeout in 1 seconds
06:58:37 main: using ipc queue 0x2d548
06:58:37 main: using ipc queue 0x2d548
06:58:37 work: received sequence 1, expected sequence 1
06:58:37 msg_dump_queue: sequence 1 in queue
06:58:37 msg_dump_queue: sequence 3 in queue
06:58:37 msg_dump: sequence 1, 'add', '/block/sda', 'block'
06:58:37 msg_exec: child [8038] created
06:58:37 running_moveto_queue: move sequence 1 [8038] to running queue '/block/sda'
06:58:37 set_timeout: set timeout in 5 seconds
06:58:37 work: received sequence 2, expected sequence 2
06:58:37 msg_dump_queue: sequence 2 in queue
06:58:37 msg_dump_queue: sequence 3 in queue
06:58:37 msg_dump: sequence 2, 'add', '/block/sdb', 'block'
06:58:37 msg_exec: child [8039] created
06:58:37 running_moveto_queue: move sequence 2 [8039] to running queue '/block/sdb'
06:58:37 msg_dump: sequence 3, 'add', '/block/sdc', 'block'
06:58:37 msg_exec: child [8040] created
06:58:37 running_moveto_queue: move sequence 3 [8040] to running queue '/block/sdc'
06:58:37 main: using ipc queue 0x2d548
06:58:37 main: using ipc queue 0x2d548
06:58:37 work: received sequence 4, expected sequence 4
06:58:37 msg_dump_queue: sequence 4 in queue
06:58:37 msg_dump: sequence 4, 'remove', '/block/sdc', 'block'
06:58:37 msg_exec: delay exec of sequence 4, [8040] already working on '/block/sdc'
06:58:37 delayed_moveto_queue: move event to delayed queue '/block/sdc'
06:58:37 msg_exec: child [8043] created
06:58:37 running_moveto_queue: move sequence 4 [8043] to running queue '/block/sdc'
06:58:37 work: received sequence 5, expected sequence 5
06:58:37 msg_dump_queue: sequence 5 in queue
06:58:37 msg_dump: sequence 5, 'remove', '/block/sdb', 'block'
06:58:37 msg_exec: delay exec of sequence 5, [8039] already working on '/block/sdb'
06:58:37 delayed_moveto_queue: move event to delayed queue '/block/sdb'
06:58:37 msg_exec: child [8044] created
06:58:37 running_moveto_queue: move sequence 5 [8044] to running queue '/block/sdb'
06:58:37 main: using ipc queue 0x2d548
06:58:37 main: using ipc queue 0x2d548
06:58:37 work: received sequence 8, expected sequence 6
06:58:37 msg_dump_queue: sequence 8 in queue
06:58:37 set_timeout: set timeout in 5 seconds
06:58:37 work: received sequence 6, expected sequence 6
06:58:37 msg_dump_queue: sequence 6 in queue
06:58:37 msg_dump_queue: sequence 8 in queue
06:58:37 msg_dump: sequence 6, 'remove', '/block/sda', 'block'
06:58:37 msg_exec: delay exec of sequence 6, [8038] already working on '/block/sda'
06:58:37 delayed_moveto_queue: move event to delayed queue '/block/sda'
06:58:37 msg_exec: child [8047] created
06:58:37 running_moveto_queue: move sequence 6 [8047] to running queue '/block/sda'
06:58:37 set_timeout: set timeout in 5 seconds
06:58:38 sig_handler: caught signal 17
06:58:38 sig_handler: exec finished, pid 8038
06:58:38 set_timeout: set timeout in 4 seconds
06:58:38 msg_dump_queue: sequence 8 in queue
06:58:38 sig_handler: caught signal 17
06:58:38 sig_handler: exec finished, pid 8039
06:58:38 set_timeout: set timeout in 4 seconds
06:58:38 msg_dump_queue: sequence 8 in queue
06:58:38 sig_handler: caught signal 17
06:58:38 sig_handler: exec finished, pid 8040
06:58:38 set_timeout: set timeout in 4 seconds
06:58:38 msg_dump_queue: sequence 8 in queue
06:58:38 sig_handler: caught signal 17
06:58:38 sig_handler: exec finished, pid 8043
06:58:38 set_timeout: set timeout in 4 seconds
06:58:38 msg_dump_queue: sequence 8 in queue
06:58:38 sig_handler: caught signal 17
06:58:38 sig_handler: exec finished, pid 8044
06:58:38 set_timeout: set timeout in 4 seconds
06:58:38 msg_dump_queue: sequence 8 in queue
06:58:38 sig_handler: caught signal 17
06:58:38 sig_handler: exec finished, pid 8047
06:58:38 set_timeout: set timeout in 4 seconds
06:58:38 msg_dump_queue: sequence 8 in queue
06:58:39 main: using ipc queue 0x2d548
06:58:39 main: using ipc queue 0x2d548
06:58:39 work: received sequence 9, expected sequence 7
06:58:39 msg_dump_queue: sequence 8 in queue
06:58:39 msg_dump_queue: sequence 9 in queue
06:58:39 set_timeout: set timeout in 3 seconds
06:58:39 work: received sequence 11, expected sequence 7
06:58:39 msg_dump_queue: sequence 8 in queue
06:58:39 msg_dump_queue: sequence 9 in queue
06:58:39 msg_dump_queue: sequence 11 in queue
06:58:39 set_timeout: set timeout in 3 seconds
06:58:39 main: using ipc queue 0x2d548
06:58:39 work: received sequence 10, expected sequence 7
06:58:39 msg_dump_queue: sequence 8 in queue
06:58:39 msg_dump_queue: sequence 9 in queue
06:58:39 msg_dump_queue: sequence 10 in queue
06:58:39 msg_dump_queue: sequence 11 in queue
06:58:39 set_timeout: set timeout in 3 seconds
06:58:39 main: using ipc queue 0x2d548
06:58:39 work: received sequence 13, expected sequence 7
06:58:39 msg_dump_queue: sequence 8 in queue
06:58:39 msg_dump_queue: sequence 9 in queue
06:58:39 msg_dump_queue: sequence 10 in queue
06:58:39 msg_dump_queue: sequence 11 in queue
06:58:39 msg_dump_queue: sequence 13 in queue
06:58:39 set_timeout: set timeout in 3 seconds
06:58:39 main: using ipc queue 0x2d548
06:58:39 work: received sequence 14, expected sequence 7
06:58:39 msg_dump_queue: sequence 8 in queue
06:58:39 msg_dump_queue: sequence 9 in queue
06:58:39 msg_dump_queue: sequence 10 in queue
06:58:39 msg_dump_queue: sequence 11 in queue
06:58:39 msg_dump_queue: sequence 13 in queue
06:58:39 msg_dump_queue: sequence 14 in queue
06:58:39 set_timeout: set timeout in 3 seconds
06:58:39 main: using ipc queue 0x2d548
06:58:39 work: received sequence 15, expected sequence 7
06:58:39 msg_dump_queue: sequence 8 in queue
06:58:39 msg_dump_queue: sequence 9 in queue
06:58:39 msg_dump_queue: sequence 10 in queue
06:58:39 msg_dump_queue: sequence 11 in queue
06:58:39 msg_dump_queue: sequence 13 in queue
06:58:39 msg_dump_queue: sequence 14 in queue
06:58:39 msg_dump_queue: sequence 15 in queue
06:58:39 set_timeout: set timeout in 3 seconds
06:58:41 main: using ipc queue 0x2d548
06:58:41 work: received sequence 12, expected sequence 7
06:58:41 msg_dump_queue: sequence 8 in queue
06:58:41 msg_dump_queue: sequence 9 in queue
06:58:41 msg_dump_queue: sequence 10 in queue
06:58:41 msg_dump_queue: sequence 11 in queue
06:58:41 msg_dump_queue: sequence 12 in queue
06:58:41 msg_dump_queue: sequence 13 in queue
06:58:41 msg_dump_queue: sequence 14 in queue
06:58:41 msg_dump_queue: sequence 15 in queue
06:58:41 set_timeout: set timeout in 1 seconds
06:58:42 sig_handler: caught signal 14
06:58:42 sig_handler: event timeout reached
06:58:42 event 8, age 5 seconds, skip event 7-7
06:58:42 msg_dump: sequence 8, 'add', '/block/sdb', 'block'
06:58:42 msg_exec: child [8057] created
06:58:42 running_moveto_queue: move sequence 8 [8057] to running queue '/block/sdb'
06:58:42 msg_dump: sequence 9, 'add', '/block/sdc', 'block'
06:58:42 msg_exec: child [8058] created
06:58:42 running_moveto_queue: move sequence 9 [8058] to running queue '/block/sdc'
06:58:42 msg_dump: sequence 10, 'remove', '/block/sdc', 'block'
06:58:42 msg_exec: delay exec of sequence 10, [8058] already working on '/block/sdc'
06:58:42 delayed_moveto_queue: move event to delayed queue '/block/sdc'
06:58:42 msg_exec: child [8059] created
06:58:42 running_moveto_queue: move sequence 10 [8059] to running queue '/block/sdc'
06:58:42 msg_dump: sequence 11, 'remove', '/block/sdb', 'block'
06:58:42 msg_exec: delay exec of sequence 11, [8057] already working on '/block/sdb'
06:58:42 delayed_moveto_queue: move event to delayed queue '/block/sdb'
06:58:42 msg_exec: child [8060] created
06:58:42 running_moveto_queue: move sequence 11 [8060] to running queue '/block/sdb'
06:58:42 msg_dump: sequence 12, 'remove', '/block/sda', 'block'
06:58:42 msg_exec: child [8061] created
06:58:42 running_moveto_queue: move sequence 12 [8061] to running queue '/block/sda'
06:58:42 msg_dump: sequence 13, 'add', '/block/sda', 'block'
06:58:42 msg_exec: delay exec of sequence 13, [8061] already working on '/block/sda'
06:58:42 delayed_moveto_queue: move event to delayed queue '/block/sda'
06:58:42 msg_exec: child [8062] created
06:58:42 running_moveto_queue: move sequence 13 [8062] to running queue '/block/sda'
06:58:42 msg_dump: sequence 14, 'add', '/block/sdb', 'block'
06:58:42 msg_exec: delay exec of sequence 14, [8057] already working on '/block/sdb'
06:58:42 delayed_moveto_queue: move event to delayed queue '/block/sdb'
06:58:42 msg_exec: child [8063] created
06:58:42 running_moveto_queue: move sequence 14 [8063] to running queue '/block/sdb'
06:58:42 msg_dump: sequence 15, 'add', '/block/sdc', 'block'
06:58:42 msg_exec: delay exec of sequence 15, [8058] already working on '/block/sdc'
06:58:42 delayed_moveto_queue: move event to delayed queue '/block/sdc'
06:58:42 msg_exec: child [8064] created
06:58:42 running_moveto_queue: move sequence 15 [8064] to running queue '/block/sdc'
06:58:43 sig_handler: caught signal 17
06:58:43 sig_handler: exec finished, pid 8057
06:58:43 sig_handler: exec finished, pid 8058
06:58:43 sig_handler: caught signal 17
06:58:43 sig_handler: exec finished, pid 8059
06:58:43 sig_handler: caught signal 17
06:58:43 sig_handler: exec finished, pid 8060
06:58:43 sig_handler: exec finished, pid 8061
06:58:43 sig_handler: caught signal 17
06:58:43 sig_handler: exec finished, pid 8062
06:58:43 sig_handler: caught signal 17
06:58:43 sig_handler: exec finished, pid 8063
06:58:43 sig_handler: caught signal 17
06:58:43 sig_handler: exec finished, pid 8064
Here is the next revision for udevd:
o Small cleanups all over the place.
o Swich to the usual linked list format "list.h".
o Better timeout handling.
We store a timestamp in in every queued event, so we don't wait longer
than the timeout specified, if the hole in the list is not shrinking.
o ignore udevd target if klibc is used
Here is the next round of udevd/udevsend:
udevsend - If the IPC message we send is not catched by a receiver we fork
the udevd daemon to process this and the following events
udevd - We reorder the events we receive and execute our current udev for
every event. If one or more events are missing, we wait
10 seconds and then go ahead in the queue.
If the queue is empty and we don't receive any event for the next
30 seconds, the daemon exits.
The next incoming event will fork the daemon again.
config - The path's to the executable are specified in udevd.h
Now they are pointing to the current directory only.
I don't like daemons hiding secrets (and mem leaks :)) inside,
so I want to try this model. It should be enough logic to get all possible
hotplug events executed in the right order.
If no event, then no daemon! So everybody should be happy :)
Here we see:
1. the daemon fork,
2. the udev work,
3. the 10 sec timeout and the skipped events,
4. the udev work,
...,
5. and the 30 sec timeout and exit.
EVENTS:
pim:/home/kay/src/udev.kay# test/udevd_test.sh
pim:/home/kay/src/udev.kay# SEQNUM=15 ./udevsend block
pim:/home/kay/src/udev.kay# SEQNUM=16 ./udevsend block
pim:/home/kay/src/udev.kay# SEQNUM=17 ./udevsend block
pim:/home/kay/src/udev.kay# SEQNUM=18 ./udevsend block
pim:/home/kay/src/udev.kay# SEQNUM=20 ./udevsend block
pim:/home/kay/src/udev.kay# SEQNUM=21 ./udevsend block
LOG:
Jan 23 15:35:35 pim udev[11795]: message is still in the ipc queue, starting daemon...
Jan 23 15:35:35 pim udev[11799]: configured rule in '/etc/udev/udev.rules' at line 19 applied, 'sda' becomes '%k-flash'
Jan 23 15:35:35 pim udev[11799]: creating device node '/udev/sda-flash'
Jan 23 15:35:35 pim udev[11800]: creating device node '/udev/sdb'
Jan 23 15:35:35 pim udev[11804]: creating device node '/udev/sdc'
Jan 23 15:35:35 pim udev[11805]: removing device node '/udev/sda-flash'
Jan 23 15:35:35 pim udev[11808]: removing device node '/udev/sdb'
Jan 23 15:35:35 pim udev[11809]: removing device node '/udev/sdc'
Jan 23 15:35:45 pim udev[11797]: timeout reached, skip events 7 - 7
Jan 23 15:35:45 pim udev[11811]: creating device node '/udev/sdb'
Jan 23 15:35:45 pim udev[11812]: creating device node '/udev/sdc'
Jan 23 15:36:01 pim udev[11797]: timeout reached, skip events 10 - 14
Jan 23 15:36:01 pim udev[11814]: creating device node '/udev/sdc'
Jan 23 15:36:04 pim udev[11816]: creating device node '/udev/sdc'
Jan 23 15:36:12 pim udev[11818]: creating device node '/udev/sdc'
Jan 23 15:36:16 pim udev[11820]: creating device node '/udev/sdc'
Jan 23 15:36:38 pim udev[11797]: timeout reached, skip events 19 - 19
Jan 23 15:36:38 pim udev[11823]: creating device node '/udev/sdc'
Jan 23 15:36:38 pim udev[11824]: creating device node '/udev/sdc'
Jan 23 15:37:08 pim udev[11797]: we have nothing to do, so daemon exits...
On Thu, Jan 22, 2004 at 01:27:45AM +0100, Kay Sievers wrote:
> On Wed, Jan 21, 2004 at 02:38:25PM +0100, Kay Sievers wrote:
> > On Thu, Jan 15, 2004 at 01:45:10PM -0800, Greg KH wrote:
> > > On Thu, Jan 15, 2004 at 10:36:25PM +0800, Ling, Xiaofeng wrote:
> > > > Hi, Greg
> > > > I wrote a simple implementation for the two pieces
> > > > of send and receive hotplug event,
> > > > use a message queue and a list for the out of order
> > > > hotplug event. It also has a timeout timer of 3 seconds.
> > > > They are now separate program. the file nseq is the test script.
> > > > Could you have a look to see wether it is feasible?
> > > > If so, I'll continue to merge with udev.
> > >
> > > Yes, very nice start. Please continue on.
> > >
> > > One minor comment, please stick with the kernel coding style when you
> > > are writing new code for udev.
> >
> > I took the code from Xiaofeng, cleaned the whitespace, renamed some bits,
> > tweaked the debugging, added the udev exec and created a patch for the current tree.
> >
> > It seems functional now, by simply executing our current udev (dirty hack).
> > It reorders the incoming events and if one is missing it delays the
> > execution of the following ones up to a maximum of 10 seconds.
> >
> > Test script is included, but you can't mix hotplug sequence numbers and
> > test script numbers, it will result in waiting for the missing numbers :)
>
> Hey, nobody want's to play with me?
> So here I'm chatting with myself :)
>
> This is the next version with signal handling for resetting the expected
> signal number. I changed the behaviour of the timeout to skip all
> missing events at once and to proceed with the next event in the queue.
>
> So it's now possible to use the test script at any time, cause it resets
> the daemon, if real hotplug event coming in later all missing nimbers will
> be skipped after a timeout of 10 seconds and the queued events are applied.
Here is the next updated updated version to apply to the lastet udev.
I've added infrastructure for getting the state of the IPC queue in the
sender and set the program to exec by the daemon. Also the magic key id
is replaced by the usual key generation by path/nr.
It looks promising, I use it on my machine and my 4in1 USB-flash-reader
connect/disconnect emits the events "randomly" but udevd is able to
reorder it and calls our normal udev in the right order.
