Date:	Mon, 02 Feb 2009 01:43:31 -0500
From:	Adam M Belay <abelay@....EDU>
To:	Robert Hancock <hancockrwd@...il.com>, bjorn.helgaas@...com
Cc:	Philippe De Muyter <phdm@...qel.be>, linux-kernel@...r.kernel.org,
	lenb@...nel.org
Subject: Re: [BUG] pnpbios breaks floppy support

Quoting Robert Hancock <hancockrwd@...il.com>:

> Philippe De Muyter wrote:
>> On Sun, Feb 01, 2009 at 03:48:49PM -0500, Adam M Belay wrote:
>>> Quoting Robert Hancock <hancockrwd@...il.com>:
>>>
>>>> Philippe De Muyter wrote:
>>>>> On Sun, Feb 01, 2009 at 01:08:33AM -0600, Robert Hancock wrote:
>>>>>> Philippe De Muyter wrote:
>>>>>>> Hello linux experts,
>>>>>>> Today I tried to upgrade a PC's kernel from 2.6.11 to 2.6.22, and
>>>>>>> I saw some strange messages when booting :
>>>>>>> 	Floppy drive(s): fd0 is 1.44M
>>>>>>> 	floppy0: Floppy io-port 0x03f2 in use
>>>>>>> Previously, I had :
>>>>>>> 	Floppy drive(s): fd0 is 1.44M
>>>>>>> 	FDC 0 is a post-1991 82077
>>>>>>> Needless to say, my floppy hardware works perfectly, and my floppy
>>>>>>> was usable with the old kernel, while the floppy is now inaccessible
>>>>>>> with the new kernel.  Even /dev/fd0 does not exist anymore.
>>>>>>> Searching for a cause to that problem, I saw the following messages
>>>>>>> before the floppy probe in the new kernel :
>>>>>>> 	PnPBIOS: Scanning system for PnP BIOS support...
>>>>>>> 	PnPBIOS: Found PnP BIOS installation structure at 0xc00fd5e0
>>>>>>> 	PnPBIOS: PnP BIOS version 1.0, entry 0xf0000:0x5ba3, dseg 0xf0000
>>>>>>> 	PnPBIOS: 17 nodes reported by PnP BIOS; 17 recorded by driver
>>>>>>> 	[...]
>>>>>>> 	pnp: 00:07: ioport range 0x3f0-0x3f1 has been reserved
>>>>>>> 	pnp: 00:07: ioport range 0x3f3-0x3f3 has been reserved
>>>>>>> 	[...]
>>>>>>> Searching the web and the outdated pnp kernel documentation, I
>>>>>>> finally found an option to add to my kernel parameters line :
>>>>>>> 	pnpbios=off
>>>>>>> Now my floppy works again, but I am not really satisfied.
>>>>>>> What do I loose with the 'pnpbios=off' option ?
>>>>>>> Isn't there a smoother option to allow pnpbios but avoiding to reserve
>>>>>>> floppy's io-ports ?
>>>>>>> Should I modify rather /drivers/block/floppy.c or /drivers/pnp/*.c
>>>>>>> to make pnpbios and floppy driver coexist peacefully ? And is there
>>>>>>> an example of such modifications for other standard peripherals ?
>>>>>> Presumably the problem is that your BIOS marks the IO ports used 
>>>>>> by the floppy controller as reserved which prevents the floppy 
>>>>>> driver from binding to them. (2.6.11 probably was before we even 
>>>>>> processed PnP reserved regions.)
>>>>>>
>>>>>> I think we now have handling for the case where the reservations 
>>>>>> overlap PCI devices, but I think it's the first I've heard of 
>>>>>> them overlapping the floppy IO ports..
>>>>> I should have added that, when started with pnpbios enabled, I 
>>>>> have found the following in /sys/devices/pnp0/ :
>>>>>
>>>>> 	$ cat 00:03/id
>>>>> 	PNP0700
>>>>> 	$ cat 00:03/resources 	state = active
>>>>> 	io 0x3f4-0x3f5
>>>>> 	io 0x3f2-0x3f2
>>>>> 	irq 6
>>>>> 	dma 2
>>>>> 	$ cat 00:03/options
>>>>> 	port 0x3f4-0x3f4, align 0x0, size 0x2, 16-bit address decoding
>>>>> 	port 0x3f2-0x3f2, align 0x0, size 0x1, 16-bit address decoding
>>>>> 	irq 6 High-Edge
>>>>> 	dma 2 8-bit compatible
>>>>>
>>>>> AFAIK, PNP0700 is the pnp id for the standard floppy disk,
>>>>> and the resources and options files describe the expected io-ports
>>>>> of the floppy disk, so this does not seem to be an error in the bios.
>>>> There's likely another resource with id of PNP0C01 or PNP0C02 
>>>> (Motherboard resources) which contains that same IO port range.
>>>>
>>> Yes, could you post the same information for 00:07 so we can start 
>>> to narrow
>>> this down?  Also having "cat /proc/ioports" couldn't hurt.
>>
>> Here it is, and that shows that you are thus both really experts in 
>> that area :
>>
>> 	$ cat 00:07/id 00:07/resources
>> 	PNP0c02
>> 	state = active
>> 	io 0x80-0x80
>> 	io 0x10-0x1f
>> 	io 0x22-0x3f
>> 	io 0x44-0x5f
>> 	io 0x90-0x9f
>> 	io 0xa2-0xbf
>> 	io 0x3f0-0x3f1
>> 	io 0x3f3-0x3f3
>> 	mem 0x100000-0xc0fffff
>> 	mem 0xfff80000-0xfff94fff
>> 	mem 0xfff98000-0xfffbffff
>> 	mem 0xfffc0000-0xffffffff
>>
>> 	$ cat /proc/ioports 	0000-001f : dma1
>> 	0020-0021 : pic1
>> 	0040-0043 : timer0
>> 	0050-0053 : timer1
>> 	0060-006f : keyboard
>> 	0070-0077 : rtc
>> 	0080-008f : dma page reg
>> 	00a0-00a1 : pic2
>> 	00c0-00df : dma2
>> 	00f0-00ff : fpu
>> 	0170-0177 : 0000:00:07.1
>> 	  0170-0177 : libata
>> 	01f0-01f7 : 0000:00:07.1
>> 	  01f0-01f7 : libata
>> 	02f8-02ff : serial
>> 	0376-0376 : 0000:00:07.1
>> 	  0376-0376 : libata
>> 	0378-037a : parport0
>> 	037b-037f : parport0
>> 	03c0-03df : vga+
>> 	03f0-03f1 : pnp 00:07
>> 	03f3-03f3 : pnp 00:07
>> 	03f6-03f6 : 0000:00:07.1
>> 	  03f6-03f6 : libata
>> 	03f8-03ff : serial
>> 	04d0-04d1 : pnp 00:11
>> 	0cf8-0cff : PCI conf1
>> 	ec00-ec7f : 0000:00:03.0
>> 	  ec00-ec7f : tulip
>> 	ecf0-ecff : 0000:00:07.1
>> 	  ecf0-ecff : libata
>>
>> Thanks for your quick answers.  Feel free to ask more info's
>
> Likely we should change things so that if a motherboard resource 
> overlaps another PnP resource then we ignore it, as obviously Windows 
> permits this behavior..
>

Ok, so it appears that we have a genuine resource conflict with the
system device (00:07).  This is because the floppy driver is requesting
four consecutive I/O ports rather than the two ports reported by the
PnPBIOS.

Here's a look at what I see on one of my ACPI boxes:

state = active
io 0x3f0-0x3f5
io 0x3f7-0x3f7
irq 6
dma 2

This should work just fine and matches well with what the floppy driver
expects.

So the next step is to determine if this issue is typical when running in
PnPBIOS mode.  I'm not convinced we need to permit PnP resources to overlap,
as it turns out to not be the issue in this case: There are no overlaps
between the system device and floppy device in Philippe's configuration.
It does suggest, however, that calling request_region() without using the
size of the resource that PnP reports can cause problems.

Bjorn, I think it might make sense to eventually have the PnP layer
register resources with request_region() instead of handling it in
device drivers.  What do you think?  In the short term we could either
skip the floppy driver's call to request_region() when using PnP detection
or have it use the length reported in the resource descriptor.  Still, I'm
curious which I/O ports the driver is actually using.

More to follow...

Thanks,
Adam

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists