Puppy Linux Discussion Forum

[greengeek]

I am trying to set up a self-booting "server" using puppy and often find that some usb sticks are not detected at boot time, but DO get detected if I unplug them after boot, and replug them ("hotplug"?)

I wondered if it might be possible to make a script that has the ability to "tickle" the usb ports in such a way that the system thinks that a new device has been plugged in - kicking off the detection process without having to physically unplug the device?

My plan would be to place this script in the /root/Startup folder so that it ran after X started. (I would probably then follow this script with some code that I got from sunburnt which automounts all drives.)

(I know that the sensitivity of usb device detection varies between versions of puppy, but I would like to try to find a piece of code that is likely to run ok on a wide range of puppies - my server is likely to be built on PupServer435 which is based on 4.3.1 or I might even use an older more barebones puppy so it would be good if the code worked on older versions, but I have also had this problem on some more recent puppies too)

[rcrsn51]

Maybe something simpler would work. Go into your auto-mount script and put a "sleep 5" before the lines that mount the USB drives.

[greengeek]

The problem is that I can't mount what has not yet been detected. I find that if the operating system itself has not detected the usb drive during the boot sequence then none of the various mount methods will work. However, if I then pull the drive out and plug it back in, an icon appears on the desktop and I can proceed to mount.

[musher0]

Hi!

Would that be a Seagate FreeAgent external USB hard drive, by any chance ? They're notorious for "sleeping on the job", and Linux has apparently no way to wake them up besides reboot. (With the exception of dpup 4.85, in my experience.)

Best.

[greengeek]

One of the devices I am mostly having this issue with is a Transcend 1TB external usb drive. I have it partitioned with one large NTFS, one small FAT32, and one Linux swap partition. ( I don't know which brand of HDD or controller chip it contains). I don't know if the non-detection is caused by the device itself or something to do with the partitioning (unfortunately I can't fiddle too much with it because it contains all my daughters uni work and media files).

However, I do have similar problems with different usb sticks - some of which have only a single FAT32 partition.

Some puppies seem very good at boot-time usb detection (Grays NOP511 is another lightweight puppy I am trialling for server use), but older puppies seem more variable. I just wanted to find a way to "simulate" a usb "hotplug" action so that the device detection routine gets triggered.

[rcrsn51]

Have a look at this. But it may not fix your problem - if a device hasn't been detected yet, you probably can't reset it.

[Edit] I found a site that claims you can restart your USB system by reloading the module.

[greengeek]

Thanks for those suggestions. The usbreset pet unfortunately did not work on the PupServer435 I am trialling because it is based on a "4" series puppy and does not understand the lsusb command.

However the rmmod/modprobe of the usb driver DID work. It allowed the missing partition to reappear. Thanks!

(I need to report back further once I have done some more testing because the cluster of symptoms I originally had seems to have changed - at the moment the one consistent symptom I have is that the usb stick is detected, but only 2 of the 3 partitions are detected - at least until I do the rmmod trick you suggested, after which all 3 partitions are back.

I need to quantify exactly what was happening across the three troublesome usb devices, two machines, and several remastered versions of Pupserver435 and NOP511. Will report back)

[greengeek]

OK, here is a script that uses the rmmod to remove the usb driver, modprobe to reload the driver, then sunburnt's second version of mnt-all code to mount all partitions.

I've tested it with PupServer435 and it is successfully getting me past my problem. I also tested it on Slacko but found that the "sleep 5" after the reload usb had to be lengthened to "sleep 10"

I will be putting this in my startup folder.

(more testing to come re the varying nature of the symptoms regarding which partitions fail to detect..)

CAUTION: this script does not attempt to first unmount any partitions that may already have been mounted, and there may be a risk that this could cause a problem, depending on when you run it. (I don't actually know how big a risk this might be. Maybe none at all. However I recommend making sure that the pfix-fsck option is active in the menu.lst entry to correct any filesystem irregularities that might result. I plan to use this script immediately after boot, and before any other programmes are activated)

[rcrsn51]

Nice work.

This module reloading technique would probably also work for people whose USB wifi adapters occasionally lock up.

[musher0]

Hello, greengeek.

Concerning

[musher0]

BTW, in case anybody needs to know how to (re)wake up a 160 Gb FreeAgent USB drive under Linux, it's here:

http://alienghic.livejournal.com/382903.html

BFN.

[greengeek]

Hi Musher0, I've just been trying inserting "umount -a" into the script before the usb reset, but getting some strange system hangs etc (when using the script on PupServer 435), so I then tried manually using umount -a in a terminal on my normal Slacko installation - and I got a lot of messages discussing the unmounting of pup-ro, pup-rw etc etc or similar messages (which looks to my untrained eye like unmounting filesystems rather than just drives...). Do you think the umount -a is safe to use or might there be a "softer" way to unmount only the drives?

Thanks for any suggestions.

[musher0]

Hello, greegeek.

> might there be a "softer" way to unmount only the drives?

Strange. I've used the unmount -a command a lot and never had a problem.

Maybe you need something like this, then:

[greengeek]

Thanks Musher0, that seems to have improved the behaviour on Slacko. The script now is:

[musher0]

Nice! The modest man forgot to include his name and the date!