Page 6 of 12

Re: version 17.1

Posted: 21 Oct 2015, 20:05
by Scooby
wyzguy wrote:ls -l /run/vdev/vdevd.log ---- Mine shows 6.5 Megabytes!!

mine is 7.6MB I know at some time I was over 10MB as I couldn't send the log
to Jude via email uncompressed anymore.

wyzguy wrote:cat /proc/sys/fs/file-nr ---- Mine shows 608 0 203935

Mine
4704 0 600139

and

Code: Select all

>  cat /proc/sys/fs/file-max
600139

I think you are reading the numbers slightly wrong
first - number off open files on machine
second - is always zero but doesn't mean you maxed out
third - max nbr of open files per machine


Now there is a max file-handles per process that is 1024 default on alphaOS

for vdevd

Code: Select all

> pidof vdevd
2043

 > ls -l /proc/2043/fd | wc -l
37


we are nowhere near max open fd per process nor max of machine as a total

This is not the root cause

Well worth a try though, didn't think about this
Thank you for the suggestion

Re: version 17.1

Posted: 21 Oct 2015, 20:36
by Scooby
OBS!!!!!! ARM Arch Rollback Machine changed to ALA Arch Linux Archive

That means that pacman won't work

The fix!

Code: Select all

sed -i -e 's/seblu.net\/a\/arm/ala.seblu.net\/repos/g' /etc/pacman.d/mirrorlist


you might have to do a pacman -Sy

I will of course update in alphaos 17.2

I have fixed a lot of issues but want to fix vdev USB problem before alphaOS 17.2

Re: version 17.1

Posted: 22 Oct 2015, 05:49
by wyzguy
Scooby,

1. lsusb -v 2>&1 | egrep -i "^bus|get|hub port [0-9]"
Developer should note any lines with "get" in them and any indented lines (like port).

2. lsusb -v 2>&1 | egrep -i "^bus|device status|iproduct"
The device status lines for controller devices are all 0x0001
On a non vdev distro (a recent puppy) which I think has either udev or eudev,
those same lines all have 0x0003. Significant?

3. Reread the red hat article again. On the second of 3 numbers "0", you looked at the
first part of the sentence (for k2.4). Look at the rest of it (for k2.6), which would apply!

4. My booted usb flashdrive is a block device.
My usb mouse is a character device. Significant?

5. The "&" on the lsusb lines above is the upper case 7 key, but you probably know that.

We are making progress. This is a learning experience for both of us.

Re: version 17.1

Posted: 22 Oct 2015, 07:18
by obarun
hi,

this soft puzzle me.

so to work properly i need to launch vdevd --once /dev before mounting rootfs in read only. then i need to launch again vdevd /dev after mounting rootfs in read write.
if i'm not launch vdevd --once /dev, vdev don't find the internet card for example.

under console, all work : usb keyboard, mouse and hotplug (if i plug an usb stick, it mounted correctly).
under x no keyboard no mouse but xorg start correctly but not detect any usb devices. So for the moment i don't care about it.

but what's realy strange is the log output, see this :

Code: Select all

[          vdev.c:0075] vdev_device_read_vdevd_instance: ERROR: vdev_read_file('/dev/metadata//dev/sda/vdev_instance') rc = -2
[          vdev.c:0156] vdev_remove_unplugged_device: ERROR: vdev_device_read_vdevd_instance('/dev/sda') rc = -2


have you see the double // in the path : metadata//dev/sda. is it normal? have you the same thing?

and this :

Code: Select all

[        action.c:0650] vdev_action_run_sync: DEBUG: run command: '/usr/lib/vdev/udev-compat.sh'
[        action.c:0653] vdev_action_run_sync: DEBUG: command env: 'VDEV_MOUNTPOINT=/dev'
[        action.c:0653] vdev_action_run_sync: DEBUG: command env: 'VDEV_ACTION=add'
[        action.c:0653] vdev_action_run_sync: DEBUG: command env: 'VDEV_PATH=UNKNOWN'
[        action.c:0653] vdev_action_run_sync: DEBUG: command env: 'VDEV_METADATA=/dev/metadata//dev/UNKNOWN'
[        action.c:0653] vdev_action_run_sync: DEBUG: command env: 'VDEV_GLOBAL_METADATA=/dev/metadata/'
[        action.c:0653] vdev_action_run_sync: DEBUG: command env: 'VDEV_CONFIG_FILE=/etc/vdev/vdevd.conf'
[        action.c:0653] vdev_action_run_sync: DEBUG: command env: 'VDEV_HELPERS=/usr/lib/vdev'
[        action.c:0653] vdev_action_run_sync: DEBUG: command env: 'VDEV_LOGFILE=/var/log/vdev/vdevd.log'

same thing here : 'VDEV_METADATA=/dev/metadata//dev/UNKNOWN'

i was take a look on subr.sh, dev-setup.sh, udev-compat.sh but i don't find the trouble.

one more if i test with a system installed on usb disk, it run correctly (under console), but if a plug an usb stick it lose my usb disk. ??????
output on console say me : I/O error

Re: version 17.1

Posted: 22 Oct 2015, 14:14
by Scooby
wyzguy wrote:1. lsusb -v 2>&1 | egrep -i "^bus|get|hub port [0-9]"
Developer should note any lines with "get" in them and any indented lines (like port).

2. lsusb -v 2>&1 | egrep -i "^bus|device status|iproduct"
The device status lines for controller devices are all 0x0001
On a non vdev distro (a recent puppy) which I think has either udev or eudev,
those same lines all have 0x0003. Significant?

WIll compare with udev results ans see if they differ

wyzguy wrote:3. Reread the red hat article again. On the second of 3 numbers "0", you looked at the
first part of the sentence (for k2.4). Look at the rest of it (for k2.6), which would apply!

In anyway it is not the cause of the libudev-compat/evdev problem in vdev

wyzguy wrote:4. My booted usb flashdrive is a block device.
My usb mouse is a character device. Significant?

Yes and it is correct.


wyzguy wrote:5. The "&" on the lsusb lines above is the upper case 7 key, but you probably know that.

Actually on my comp it's the upper case 6 key

wyzguy wrote: We are making progress. This is a learning experience for both of us.

yeah but I hope I don't have to learn how to debug Xorg server to solve it

Re: version 17.1

Posted: 22 Oct 2015, 14:30
by Scooby
obarun wrote:this soft puzzle me.

yeah me too

obarun wrote:so to work properly i need to launch vdevd --once /dev before mounting rootfs in read only. then i need to launch again vdevd /dev after mounting rootfs in read write.
if i'm not launch vdevd --once /dev, vdev don't find the internet card for example.

What do you have in /dev before running vdev --once ?

obarun wrote:under x no keyboard no mouse but xorg start correctly but not detect any usb devices. So for the moment i don't care about it.

Yep happens every time I compile a new vdev version.
I know some of the times I did wrong /usb/lib/libudev links
you might try adding some symlink versions here for intance libudev.so.1.5.2

obarun wrote:but what's realy strange is the log output, see this :

Code: Select all

[          vdev.c:0075] vdev_device_read_vdevd_instance: ERROR: vdev_read_file('/dev/metadata//dev/sda/vdev_instance') rc = -2
[          vdev.c:0156] vdev_remove_unplugged_device: ERROR: vdev_device_read_vdevd_instance('/dev/sda') rc = -2


sda something you unplugged at console???

obarun wrote:have you see the double // in the path : metadata//dev/sda. is it normal? have you the same thing?

Yes and I asked Jude about it cause I thought there was like a variable that was empty by misstake there
His answer was that it was a "defensive" / if it was missing from another variable, like for robustness.

It is OK

for instance try something like ls //tmp and it works, right

obarun wrote:one more if i test with a system installed on usb disk, it run correctly (under console), but if a plug an usb stick it lose my usb disk. ??????
output on console say me : I/O error

Strange, if you can repeat the problem consistently you should report it at vdev github

Re: version 17.1

Posted: 23 Oct 2015, 08:24
by obarun
i use a classic archlinux initramfs. a hook start udev make stuff and kill the daemon; then the control is passed to runit for PID 1.
i try to make a hook for vdev but for the moment it's not ready. problem with pid in the initramfs, vdevd don't create pid and log. i have modified for this version the classic init script but not work. i need more test and work on it.

anyway, after the initramfs i try to launch vdevd as explained on previous post

What do you have in /dev before running vdev --once ?


a complete aborescence without input directory on it : this is normal.

Yep happens every time I compile a new vdev version.
I know some of the times I did wrong /usb/lib/libudev links
you might try adding some symlink versions here for intance libudev.so.1.5.2


my pkgbuild do that automaticly (pkgbuild on my github is not updated for the moment). i use libudev.so.1.6.3. i don't think that is a version problem.

sda something you unplugged at console???


not. it's just to line but i have more with other device

Yes and I asked Jude about it cause I thought there was like a variable that was empty by misstake there
His answer was that it was a "defensive" / if it was missing from another variable, like for robustness.


Well ok for that thank

Strange, if you can repeat the problem consistently you should report it at vdev github


yep consistently (maked 4 try). i will do a report bug but not for the moment.

So i have tried a little more deeply your iso. Are you using vdev-compat? if i look on your /usr/share/pkgconfig i don't see libudev.pc but udev.pc. it's normal? i go to make a test with it. The keyboard with your iso is finded but not the mouse.
onemore i don't have /sys/kernel/uevent_helper on my aborescence when i test vdevd. i don't understand why. Do you use vdevfs?

you use kickshaw, i not knew this, thank for that. i was made a pkgbuild for it available on my github and repo

if i can make a critizism. launch dhcpcd (not a daemon) in your init to provide a direct acces on internet booting in text mode making a resolv.conf. Add more server on mirrorlist.
Ram used under X is 200M. Can you reduce this? it's a lot given the started daemons.

are you using /etc/init.d/vdev file to launch the daemon? if it's the case , try to reduce your booting time by changing the sleep time parameter on this line

Code: Select all

if [ -d /sys/class/mem/null -a ! -L /sys/class/mem/null ] || \
   [ -e /sys/block -a ! -e /sys/class/block ]; then
  log_warning_msg "CONFIG_SYSFS_DEPRECATED must not be selected"
  log_warning_msg "Booting will continue in 30 seconds but many things will be broken"
  sleep 30
fi


finaly i very impressive by the concept of your iso. keep up your good work :)

Re: version 17.1

Posted: 23 Oct 2015, 18:27
by Scooby
obarun wrote:So i have tried a little more deeply your iso. Are you using vdev-compat?

I use vdev's libudev-compat, yes.

obarun wrote: if i look on your /usr/share/pkgconfig i don't see libudev.pc but udev.pc. it's normal? i go to make a test with it. The keyboard with your iso is finded but not the mouse.

Nah there should be a libudev.pc. In fact in my latest version(not uploaded) there is.
Will release a fixed iso as soon as USB mouse issue with vdev is solved.


obarun wrote:onemore i don't have /sys/kernel/uevent_helper on my aborescence when i test vdevd. i don't understand why.

What is that? uevent_helper?


obarun wrote:Do you use vdevfs?

No


obarun wrote:you use kickshaw, i not knew this, thank for that. i was made a pkgbuild for it available on my github and repo

your'e welcome

obarun wrote:if i can make a critizism. launch dhcpcd (not a daemon) in your init to provide a direct acces on internet booting in text mode making a resolv.conf.

Constructive criticism is good.

I will have to think about your suggestion, but I will consider it.

obarun wrote:Add more server on mirrorlist.

Are you speaking about /etc/pacman.d/mirrorlist ?

alphaos uses arch linux archive of a specific date so only seblu.net to my knowledge availible as mirror.
https://wiki.archlinux.org/index.php/Arch_Linux_Archive


obarun wrote:Ram used under X is 200M. Can you reduce this? it's a lot given the started daemons.

Simargl was a master at this.
Sure things could be done but for me personally it doesn't matter.

I have 6gigs of RAM and I never max it out - I use no swap file.
I boot with toram boot parameter that copies all to RAM, thats how I like it.

A special reason why you want to lower the RAM usage?

obarun wrote:are you using /etc/init.d/vdev file to launch the daemon?

yes

obarun wrote:if it's the case , try to reduce your booting time by changing the sleep time parameter on this line

Code: Select all

if [ -d /sys/class/mem/null -a ! -L /sys/class/mem/null ] || \
   [ -e /sys/block -a ! -e /sys/class/block ]; then
  log_warning_msg "CONFIG_SYSFS_DEPRECATED must not be selected"
  log_warning_msg "Booting will continue in 30 seconds but many things will be broken"
  sleep 30
fi


When I check the conditional statements they evaluate so the sleep doesn't run on my system
My kernel does not use CONFIG_SYSFS_DEPRECATED

obarun wrote:finaly i very impressive by the concept of your iso. keep up your good work :)


Thank you, thank you, tank you, you are far to kind.

It is good with feedback

Re: version 17.1

Posted: 23 Oct 2015, 18:29
by Scooby
wyzguy wrote:2. lsusb -v 2>&1 | egrep -i "^bus|device status|iproduct"
The device status lines for controller devices are all 0x0001
On a non vdev distro (a recent puppy) which I think has either udev or eudev,
those same lines all have 0x0003. Significant?


wyzguy I now crosschecked this with udev and got the same "device status" as with vdev on my machine.
I believe this is normal.

I test this on alphaOS 16 where I can switch between udev and vdev.

I cannot explain why you get a different result with puppy.

Re: version 17.1

Posted: 23 Oct 2015, 19:51
by obarun
What is that? uevent_helper?

i don't known exactly but it seems to be important if i refer at hook example file given by jcnelson

Code: Select all

if [ -w /sys/kernel/uevent_helper ]; then
   echo > /sys/kernel/uevent_helper
fi


obarun wrote:Add more server on mirrorlist.

Are you speaking about /etc/pacman.d/mirrorlist ?

yep
alphaos uses arch linux archive of a specific date so only seblu.net to my knowledge availible as mirror.
https://wiki.archlinux.org/index.php/Arch_Linux_Archive

ok i understand better

Simargl was a master at this.
Sure things could be done but for me personally it doesn't matter.

I have 6gigs of RAM and I never max it out - I use no swap file.
I boot with toram boot parameter that copies all to RAM, thats how I like it.

A special reason why you want to lower the RAM usage?

no just for optimize the configuration


Thank you, thank you, tank you, you are far to kind.

It is good with feedback

i known the problem, same thing for me. but in french : "pas de nouvelles, bonnes nouvelles". it means : not news, so good news! no problem on it lol

Well, i progress on the hook for vdevd and i hope it will be available soon. i would like to eradicate udev from my sytem. i learned on lot today about kdbus and i have a bad feeling about this. Only gentoo provide an alternative for udev for the moment and it's not a good way to have only a few developpers working on an important soft like that. i hope a lot about vdev but for the moment for me it doesn't work.