Re: hardrive disconnected when pluging dvd burner

[Stefan R. <st...@s5...>]

I wrote:
> damien benoist wrote:
>> I have a dvd burner and a hardrive pluged on a hub.
>> When I turn on the dvd burner, if the hardrive is already turned on and
>> detected, it seems to be "disconnected" and "reconnected".
>> If there is a mounted partition on the hard drive, the acces to it is 
>> lost,
>> it has to be unmounted (fsck'ed) and remounted.
>> It seems to be a constant behaviour (I've never seen it not happen).
> 
> Astonishing.  I have two different kinds of 6-port hubs (a rather old 
> one and a recently bought one; they are most certainly based on the same 
> TI chip though) and an old portable CD-RW --- and this combined with 
> apparently any host controller and disk behaves exactly as you 
> described.  And I never noticed it.
> 
>> XP seems to not have the problem on the same hardware.
> 
> I presume they handle temporary connection loss better.  OS X definitely 
> does; there it is even possible to plug a disk out and back in while it 
> is mounted if it's done reasonably quickly.  That's handy if one has too 
> few ports and needs to rearrange the daisy chain.
> 
> And there are two more scenarios in which the kernel should handle brief 
> connection loss but doesn't yet:
>   - There seems to be a bug in some 1394b PHYs which cause the bus
>     breaking down for a moment during heavy traffic:
>     http://bugzilla.kernel.org/show_bug.cgi?id=10307
>   - I have a 1394b disk which continues to work as a repeater when
>     switched off.  This repeater function works by drawing bus power
>     for the PHY.  Alas the transition from self-power to bus power or
>     back causes the chain to break for a blink of an eye.
> 
> The "classic" ieee1394 Linux drivers are affected by all these problems 
> except the last one.  The new firewire drivers are affected by all of 
> them including the last one, because they monitor the related bus events 
> more precisely than ieee1394.

Hmm, the last issue with the temporarily disable repeater function of my 
1394b disk currently also affects the ieee1394 drivers.  I'm not so sure 
anymore if they were really able to cover up this gap in older versions.

...
> Thanks for the detailed information.  I can reproduce the exact same 
> results here.  I will try to fix it, but it may take me a while.  I am 
> now tracking it as http://bugzilla.kernel.org/show_bug.cgi?id=11349 .

I got some results; quicker than I expected.  I will post three patches 
to linux1394-devel (and Cc you) which fix the "plug cdrw into 6-port 
hub" for me.  The 1st patch is only about a side-issue which I stumbled 
upon when reading the relevant code.  The 2nd patch gets PC + hub + disk 
+ CD-RW going.  The 3rd patch is for larger buses with several disks.

These patches do not fix though:
   - The power transition issue of my 1394b disk as repeater,
   - connection loss if I plug a second bus-powered 6-port hub into a bus
     consisting of PC--hub--disk.
But that shouldn't matter to you since you use only one hub, and that 
one is self-powered according to our previous mail exchange.

Nevertheless, the behaviour of the PHY of your CD-RW may be subtly 
different from that of my CD-RW, hence we will have to see how your 
testing will go.
-- 
Stefan Richter
-=====-==--- =--- =----
http://arcgraph.de/sr/