[SeaBIOS] Re: Floppy drive configuration

[Eduardo Batalha]

Hi Petr,

Given the comments from Kevin, I have reverted my seabios floppy.c file 
to what it was before my changes.
Seabios is not the right place to configure the CMOS_FLOPPY_DRIVE_TYPE 
in the rtc, I think that role is probably better for coreboot itself.
So, just for completeness sake, even if it's a bit out of scope for this 
mailing list, here is what I did:
I changed the coreboot cmos.layout and cmos.default files for my Asus 
P5Q board like this
cmos.layout -
"
.
.
.
entries

# -----------------------------------------------------------------
0    120    r    0    reserved_memory

# -----------------------------------------------------------------
# FLOPPY
132    4    e    3    floppyA
128    4    e    3    floppyB

# -----------------------------------------------------------------
# RTC_BOOT_BYTE (coreboot hardcoded)
384    1    e    4    boot_option
388    4    h    0    reboot_counter
.
.
.
.

enumerations

#ID    value    text
1    0    Disable
1    1    Enable
2    0    Enable
2    1    Disable
3    0    None
3    1    360KB
3    2    1.2MB
3    3    720KB
3    4    1.44MB
3    5    2.88MB
3    6    160KB
3    7    180KB
3    8    320KB
4    0    Fallback
4    1    Normal
6    0    Emergency
6    1    Alert
.
.
.
"

cmos.default -
"
boot_option=Fallback
floppyA=None
floppyB=None
debug_level=Debug
power_on_after_fail=Disable
nmi=Enable
sata_mode=AHCI
"

This way, the floppy type for both Drives A and B can be configured in 
the secondary payload nvramcui.

The cbmem log shows
"[WARN ]  RTC: Checksum invalid"
but that does not seem to cause any problems in the few tests I did so far.

This is a hack that works for my specific case and may work on other 
boards with a floppy interface. It needs a lot more testing...

Regards,
Ed

Eduardo Batalha via SeaBIOS wrote:
> Hi,
>
> I had some old files saved in diskettes that I really needed to see, 
> so I bought a floppy drive and installed it in my coreboot based 
> computer, so, no, It's not a retro project. :)
>
> Regarding the AHCI CDROM issues, I really don't know. I did not have 
> problems with it so I did not look at that code at all.
>
> Going back to the floppy configuration issue I think I need to explain 
> my logic a bit better.
> So... Seabios does not provide us with a fancy interface to configure 
> the floppy drive type. It also does not assume anything. For real 
> hardware, (not qemu,) it is completely dependent on the etc/floppy 
> "file" being present in CBFS. The reason it cannot depend on the cmos 
> value is because when you remove the RTC battery, the CMOS memory gets 
> filled up with random values. coreboot, which runs before seabios has 
> a mechanism to fix this, but ignores the floppy type byte, leaving it 
> with a random value.
>
> My idea below, with my code, is not to use the CMOS value if there are 
> no etc/floppy files but instead to update the cmos value 
> unconditionally whenever etc/floppy files exist.
> romfile_loadint returns the second argument if the file does not 
> exist, i.e. 0. so the if condition is correct for the logic I 
> explained above.
>
> On a second thought, I now think the condition "if (type0 || type 1)" 
> should not even exist. The best solution is to have the 
> CMOS_FLOPPY_DRIVE_TYPE always updated on every boot. This way, if 
> there are no etc/floppy files in CBFS the value of the byte would be 
> zero, indicating to the operating system that there are no floppy 
> drives installed.
>
> Cheers,
> Ed
>
> On Wednesday, 3 April 2024 at 03:58:40 BST, Petr Cvek 
>  wrote:
>
> Thanks for info,
>
> So I guess yours ICH10 AHCI controller can probably withstand an NULL 
> pointer access in hardware. Unlike my ICH7, which had problems with 
> the same situation, as described here:
>
> https://www.mail-archive.com/[e-mail]/msg12950.html
> > > if (type0 || type 1)
> > >             rtc_write(CMOS_FLOPPY_DRIVE_TYPE, type0 <<4 | type1);
>
> Oh BTW I was looking how/if the code meanwhile changed and I think it 
> should be:
>
> if (!type0 || !type 1)
>         rtc_write(CMOS_FLOPPY_DRIVE_TYPE, type0 <<4 | type1);
>
> or something as in "if none romfile_loadint() set type, use CMOS values".
>
> But on the other hand, if CMOS load is allowed even on a real 
> hardware, then the entire "if (CONFIG_QEMU) .. else ..." block can be 
> optimized out.
>
> BTW^2 I completely forgot to ask. What are you trying to use the 
> floppy for, some retro project?
>
> Best regards,
> Petr
>
> Dne 03. 04. 24 v 0:45 Eduardo Batalha napsal(a):
> > Hi,
> >
> > Yes, booting from a SATA DVD drive with the controller configured as
>
> AHCI worked fine for me.
> > Cheers,
> > Ed
> >
> > On Tuesday, 2 April 2024 at 02:19:42 BST, Petr Cvek
>
> <[e-mail] > wrote:
> > And I agree ;-)
> >
> > it worked on Kontron 986LCD, but I don't have time to play with it
>
> (and I upgraded to a ryzen board). But actually I'm planing to use all 
> semi-standard CMOS addresses for legacy/retro boards.
> > BTW does a boot from SATA CD/DVD works for you with seabios? (native
>
> IDE CD/DVD on PATA is in question too)
> > Petr
> >
> > Dne 02. 04. 24 v 2:59 Eduardo Batalha via SeaBIOS napsal(a):
> > > Hi all,
> > >
> > > I own an ASUS P5Q and had the same problem that Petr Cvek mentioned
>
> in his previous emails.
> > > I was configuring the floppy type using the the etc/floppy0 value
>
> in cbfs but took me quite a while to figure out that why linux was not 
> showing it.
> > > I haven't tested it myself yet, but I think the solution to the
>
> floppy configuration issue may be to continue using etc/floppyX and 
> change these lines in floppy.c from
> > > } else {
> > >         u8 type = romfile_loadint("etc/floppy0", 0);
> > >         if (type)
> > >             addFloppy(0, type);
> > >         type = romfile_loadint("etc/floppy1", 0);
> > >         if (type)
> > >             addFloppy(1, type);
> > >     }
> > >
> > > to
> > >
> > > } else {
> > >         u8 type0 = romfile_loadint("etc/floppy0", 0);
> > >         if (type0)
> > >             addFloppy(0, type);
> > >         u8 type1 = romfile_loadint("etc/floppy1", 0);
> > >         if (type1)
> > >             addFloppy(1, type1);
> > >         if (type0 || type 1)
> > >             rtc_write(CMOS_FLOPPY_DRIVE_TYPE, type0 <<4 | type1);
> > >     }
> > >
> > > Quoting from osdev.org :
> > > "The first 14 CMOS registers access and control the Real-Time
>
> Clock. In fact, the only truly useful registers remaining in CMOS are 
> the Real-Time Clock registers, and register 0x10. All other registers 
> in CMOS are almost entirely obsolete (or are not standardized), and 
> are therefore useless. "
> > > So, register 0x10, the floppy drive type, is, conversely, pretty
>
> much standardised...
> > > Hope this makes sense...
> > >
> > > Cheers,
> > > Ed
> > > _______________________________________________
> > > SeaBIOS mailing list -- [e-mail]
>
> > > To unsubscribe send an email to [e-mail]
>
> > SeaBIOS mailing list -- [e-mail]
>
> > To unsubscribe send an email to [e-mail]
>
>  
> _______________________________________________
> SeaBIOS mailing list -- [e-mail] 
> To unsubscribe send an email to [e-mail] 
>
> SeaBIOS mailing list -- [e-mail]
> To unsubscribe send an email to [e-mail]