floppy support

[Hans de Goede]

Hi,

On 08/30/2011 10:22 PM, Chris Adams wrote:
> Once upon a time, Simo Sorce<simo at redhat.com>  said:
>> They do not 'hang', they just take longer to boot, sometimes a lot
>> longer.
>
> How much longer?

Much much longer, when I was still on the anaconda team we had
numerous bug reports about this (esp during RHEL-6 testing), in many
cases people filed bugs with a description along the lines of:
"Installation DVD does not boot", because it took so long they thought
the install was just hanging forever.

> How many such machines?

I've no hard numbers but enough to generate numerous bug reports
during the non public testing phase of RHEL-6 alone.

Also see:
https://bugzilla.redhat.com/show_bug.cgi?id=565693
https://bugzilla.redhat.com/show_bug.cgi?id=587909
https://bugzilla.redhat.com/show_bug.cgi?id=682426

http://lists.fedoraproject.org/pipermail/kernel/2010-April/002394.html
http://lists.fedoraproject.org/pipermail/kernel/2010-April/002420.html

http://docs.redhat.com/docs/en-US/Red_Hat_Enterprise_Linux/6/html/Technical_Notes/index.html
And search for "IBM ThinkPad T43 notebook"

Note this is in no way an exhaustive list of all bugs about this,
just something a really really quick search turned up.

> Again, I've booted systems
> without floppy drives but with floppy support loaded, and I haven't seen
> any significant hang.

Yes the hang is not guaranteed to be there, or to take very long,
but often it is there, and sometimes it takes very long (which is
the real problematic case). If you're really interested in seeing this fixed
go talk to the kernel guys:

anaconda loads the floppy driver by default when booting of the install
DVD, because of driverdisk support, thus the anaconda team has been getting
its share of bug reports wrt this. But AFAIK there are no such issues
in RHEL-5, where we auto-load the floppy driver too, so something
has changed in the kernel causing the hang when no drive is attached to the
controller. I could swear I filed a bug against the kernel about this
regression, but I cannot find it.

> Leaving known-working hardware unusable at install is just rude and
> irritating when it is needed.  There should be good justification, not
> just "a bunch of developers don't use it anymore, so we don't think
> anybody else needs it".

Is it delays bootup by up to 10-30 minutes on various modern systems
a good enough justification?

Regards,

Hans