usb floppy delays in /sys

[David Zeuthen]

On Thu, 2004-09-16 at 17:58 +0100, Luke Kenneth Casson Leighton wrote:
> > Cool, you got a patch for this? It works out of the box for me with my
> > Lacie USB Floppy Drive so I can't really reproduce the problem.
>  
>  i'm using selinux [on debian.  with udev. don't ask] so
>  there's a significant delay of about 1/3 to 1/2 a second on device
>  inode creation as it is.
> 

I presume you mean device node creation.

>  so it's a more a matter of you _would_ possibly come across it -
>  on a particularly busy computer - and have things inexplicably fail.
> 

Oh, but we already take this into account! In fact we reliably handle
(or at least we will 100% once udev passes the SEQNUM) much larger
delays - see this message discussing the delays incurred by hotplugging
an entire USB hub worth of devices for some background

 http://freedesktop.org/pipermail/hal/2004-August/000853.html

>  what i've added [above] falls into the "ItWorksForMe(tm)" category
>  and it was very easy do add: track down the debug message "d is NULL!"
>  and add in the sleep(1).
> 
>  i did the same thing for the hacked-up version of usbmount because that
>  shellscript suffered from the same problem [it uses /usr/bin/sg_map
>  and also hunts through /sys only to find that the entry isn't there
>  but is 5 seconds later].
> 
> 
>  regarding a patch: to hal_hotplug.c?
> 
>  i wouldn't know where to start... okay... how would i determine the
>  correct sysfs file type?
> 

So, basically, yeah, if you can remove your "ItWorksForMe(tm)" patch and
send the output of 'hald --daemon=no --verbose=yes' from when you plugin
the drive I'm sure we can easily track down what sysfs files makes this
fail. Then we can patch hal_hotplug.c to cope with that and all should
be good.

Much thanks,
David
_______________________________________________
hal mailing list
hal at freedesktop.org
http://freedesktop.org/mailman/listinfo/hal