etherboot-developers — Discussion of developer issues

Re: [Etherboot-developers] vg100

[<ke...@us...>]

>Is the hp100 (HP anylan VG100) supported by etherboot?

Feel free to write a driver..

RE: Re: [Etherboot-developers] 5.1 will not boot on 486

[Timothy L. <tl...@ro...>]

> Not unless CONSOLE_SERIAL is enabled in the current cvs.  I had
checked
> out a clean cvs prior to committing the isapnp changes.  There are
serial
> ports on the pc, so even if it was, it sould work.  I can check it out
> anyway.

I actually commented out everything in main() before the printf and it
failed in the same place.

Tim

[Etherboot-developers] vg100

[Pieter P. <pie...@pp...>]

Is the hp100 (HP anylan VG100) supported by etherboot?

Re: Re: [Etherboot-developers] 5.1 will not boot on 486

[Timothy L. <tl...@ro...>]

> The R is the first character printed by main(). I have seen this before
> where it's also trying to write to the serial port at the same and
> there's no serial port on the machine. However CONSOLE_SERIAL is not
> enabled by default. It's not on your build by any chance is it?

Not unless CONSOLE_SERIAL is enabled in the current cvs.  I had checked out a clean cvs prior to committing the isapnp changes.  There are serial ports on the pc, so even if it was, it sould work.  I can check it out anyway.

Tim

Re: [Etherboot-developers] 5.1 will not boot on 486

[<ke...@us...>]

>Yes, from a clean cvs I did a:
>
>cd etherboot/etherboot-5.1/src
>make clean
>make Roms
>make bin/3c509.fd0
>
>That image had the error.  However, while developing the sundance driver, I w
>as using the same process but instead booting a Pentium 200.  The process wor
>ked fine on the Pentium 200.

The R is the first character printed by main(). I have seen this before
where it's also trying to write to the serial port at the same and
there's no serial port on the machine. However CONSOLE_SERIAL is not
enabled by default. It's not on your build by any chance is it?

Re: Re: [Etherboot-developers] 5.1 will not boot on 486

[Timothy L. <tl...@ro...>]

> By 5.1 I assume you mean what's in CVS. If not you should get updated.
> It's probably not anything to do with your 486 but the floppy loader.
> (Unless you have seen that it works on a non-486.) What are you loading
> .dsk or .zdsk?

Yes, from a clean cvs I did a:

cd etherboot/etherboot-5.1/src
make clean
make Roms
make bin/3c509.fd0

That image had the error.  However, while developing the sundance driver, I was using the same process but instead booting a Pentium 200.  The process worked fine on the Pentium 200.

Tim

Re: [Etherboot-developers] 5.1 will not boot on 486

[<ke...@us...>]

>I reported this issue some time ago and it looks like it is still there.  I t
>ried both the 3c515 driver and the 3c509 driver, but neither even get to the 
>stage of looking for the card.  They both freeze hard at:
>
>.000041.0000 done
>R
>
>It PC in question is a Prolinea 4/25 with an Evergreen 133 processor upgrade.

By 5.1 I assume you mean what's in CVS. If not you should get updated.
It's probably not anything to do with your 486 but the floppy loader.
(Unless you have seen that it works on a non-486.) What are you loading
.dsk or .zdsk?

[Etherboot-developers] 5.1 will not boot on 486

[Timothy L. <tl...@ro...>]

Hi

I reported this issue some time ago and it looks like it is still there.  I tried both the 3c515 driver and the 3c509 driver, but neither even get to the stage of looking for the card.  They both freeze hard at:

.000041.0000 done
R

It PC in question is a Prolinea 4/25 with an Evergreen 133 processor upgrade.

I realize that it is old and somewhat unique, unfortunately, it is also where I worked on the sundance driver and the only PC where I can do any work on the ISAPNP support for the 3c515.

Any thoughts on how to track down this issue?

Tim

FROM: Timothy LeggeDATE: 08/26/2002 18:53:23SUBJECT: RE:  [Etherboot-developers] [COMMIT] 5.1.2+ boot from disk. Hi

No time for extensive testing, but just a couple of things.  I get an error:

segment [00092800,00094000) does not fit in any memory region

I am assuming that this is because the ltsp kernel is not tagged with the
correct version of mknbi.  Is that correct?

Also, on my old 486, the 3c515 driver with isa_pnp defined stops at:

.000041.0000 done
R

I will try another card (3c509) in that PC and I will try the 3c515 in
another PC when I get a chance.

Tim

Re: [Etherboot-developers] 5.1.4/5: more relocation fixes

[<ebi...@ln...>]

ke...@us... writes:

> Checked in a bunch of possible fixes for relocation for natsemi, fa311,
> and davicom drivers.
> 
> The lance based drivers look hard to fix because it has only a 24-bit
> address register, it wasn't meant to do DMA above 16MB.

Hmm.  I wonder if the pci version have that problem, or if it is just
the old ISA versions.  In which case it might make sense to split the lance
driver.

At the very least fixing it would allow etherboot to move around a little
bit in memory....

> Guess I better ready a 5.1.5 release :-|.

That is the pain with development releases. 

Eric

[Etherboot-developers] [COMMIT] 5.1.5-pre2....

[<ebi...@ln...>]

With this checking, except for moving some declarations into header
files, and possibly figuring out how to generate an efi pci option
rom.  The ia64 efi support is pretty much complete.

- Rename CONSOLE_CRT CONSOLE_FIRMWARE and make it work under efi
- Move all program defines into CONFIG for better cross compiling support
- Rename $(GCC) to $(HOST_CC) to make it clearer the difference in the to compilers is
- In genrules.pl remove the prototype for __gendep so perl won't complain that it can't use it
- Final fixes to get the efi undi_nii driver working.


Eric

[Etherboot-developers] 5.1.4/5: more relocation fixes

[<ke...@us...>]

Checked in a bunch of possible fixes for relocation for natsemi, fa311,
and davicom drivers.

The lance based drivers look hard to fix because it has only a 24-bit
address register, it wasn't meant to do DMA above 16MB.

Guess I better ready a 5.1.5 release :-|.

[Etherboot-developers] 5.1.4/5: patch for sis900.c

[<ke...@us...>]

And hopefully this makes the sis900 driver work with relocation.

Index: sis900.c
===================================================================
RCS file: /cvsroot/etherboot/etherboot/etherboot-5.1/src/drivers/net/sis900.c,v
retrieving revision 1.2
diff -u -r1.2 sis900.c
--- sis900.c	27 Dec 2002 12:44:36 -0000	1.2
+++ sis900.c	9 Jan 2003 16:34:27 -0000
@@ -651,10 +651,10 @@
 {
     txd.link   = (u32) 0;
     txd.cmdsts = (u32) 0;
-    txd.bufptr = (u32) &txb[0];
+    txd.bufptr = virt_to_bus(&txb[0]);
 
     /* load Transmit Descriptor Register */
-    outl((u32) &txd, ioaddr + txdp); 
+    outl(virt_to_bus(&txd), ioaddr + txdp); 
     if (sis900_debug > 0)
         printf("sis900_init_txd: TX descriptor register loaded with: %X\n", 
                inl(ioaddr + txdp));
@@ -679,16 +679,16 @@
 
     /* init RX descriptor */
     for (i = 0; i < NUM_RX_DESC; i++) {
-        rxd[i].link   = (i+1 < NUM_RX_DESC) ? (u32) &rxd[i+1] : (u32) &rxd[0];
+        rxd[i].link   = virt_to_bus((i+1 < NUM_RX_DESC) ? &rxd[i+1] : &rxd[0]);
         rxd[i].cmdsts = (u32) RX_BUF_SIZE;
-        rxd[i].bufptr = (u32) &rxb[i*RX_BUF_SIZE];
+        rxd[i].bufptr = virt_to_bus(&rxb[i*RX_BUF_SIZE]);
         if (sis900_debug > 0)
             printf("sis900_init_rxd: rxd[%d]=%X link=%X cmdsts=%X bufptr=%X\n", 
                    i, &rxd[i], rxd[i].link, rxd[i].cmdsts, rxd[i].bufptr);
     }
 
     /* load Receive Descriptor Register */
-    outl((u32) &rxd[0], ioaddr + rxdp);
+    outl(virt_to_bus(&rxd[0]), ioaddr + rxdp);
 
     if (sis900_debug > 0)
         printf("sis900_init_rxd: RX descriptor register loaded with: %X\n", 
@@ -1030,7 +1030,7 @@
     outl(TxDIS | inl(ioaddr + cr), ioaddr + cr);
 
     /* load Transmit Descriptor Register */
-    outl((u32) &txd, ioaddr + txdp); 
+    outl(virt_to_bus(&txd), ioaddr + txdp); 
     if (sis900_debug > 1)
         printf("sis900_transmit: TX descriptor register loaded with: %X\n", 
                inl(ioaddr + txdp));

Re: [Etherboot-developers] ISAPNP Separation Committed

[<ke...@us...>]

>1) Replacing 3c515_isapnp.h with isapnp.h
>2) Moving the ISAPNP code to isapnp.c
>
>Ken, it would be worth reviewing my Makefile changes. 

Looks good, compiles ok. Have checked in a few other changes to the 5.0
tree so you might want to cvs update to sync.

[Etherboot-developers] 5.1.4/5: patch for tulip.c

[<ke...@us...>]

Posting this because the tulip is a widely used chip and any
confirmation that this patch works will allow us to tick off this driver
as working for 5.1.x.

Index: tulip.c
===================================================================
RCS file: /cvsroot/etherboot/etherboot/etherboot-5.1/src/drivers/net/tulip.c,v
retrieving revision 1.2
diff -u -r1.2 tulip.c
--- tulip.c	27 Dec 2002 12:44:36 -0000	1.2
+++ tulip.c	9 Jan 2003 12:44:01 -0000
@@ -1014,8 +1014,8 @@
     }
 
     /* Point to rx and tx descriptors */
-    outl((unsigned long)&rx_ring[0], ioaddr + CSR3);
-    outl((unsigned long)&tx_ring[0], ioaddr + CSR4);
+    outl(virt_to_le32desc(&rx_ring[0]), ioaddr + CSR3);
+    outl(virt_to_le32desc(&tx_ring[0]), ioaddr + CSR4);
 
     init_media(nic);
 

[Etherboot-developers] ISAPNP Separation Committed

[Timothy L. <tl...@ro...>]

Hi

I committed the changes to separate the ISAPNP code from the 3c515.c file.  Changes include:

1) Replacing 3c515_isapnp.h with isapnp.h
2) Moving the ISAPNP code to isapnp.c

Ken, it would be worth reviewing my Makefile changes. 

Tim

Re: [Etherboot-developers] gendep() called too early?

[<ebi...@ln...>]

ke...@us... writes:

> /usr/bin/perl ./genrules.pl i386 NIC  core/serial.c core/main.c
> core/pci.c core/osloader.c core/nfs.c core/misc.c core/config.c
> core/relocate.c core/heap.c drivers/disk/floppy.c core/nic.c core/disk.c
> core/timer.c core/proto_slam.c core/proto_tftm.c
> arch/i386/prefix/boot1a.s arch/i386/prefix/unhuf.S
> arch/i386/prefix/unnrv2b.S arch/i386/firmware/pcbios/memsizes.c
> arch/i386/firmware/pcbios/pcbios.S arch/i386/prefix/liloprefix.S
> arch/i386/prefix/elfprefix.S arch/i386/prefix/comprefix.S
> arch/i386/prefix/pxeprefix.S arch/i386/prefix/loader.S
> arch/i386/core/start32.S arch/i386/core/start16.S arch/i386/core/pxe.c
> arch/i386/core/pci_io.c arch/i386/core/i386_timer.c arch/i386/core/elf.c
> > bin/Roms
> main::__gendep() called too early to check prototype at ./genrules.pl line 48.
> Driver file arch/ia64/drivers/net/undi_nii.c not found, skipping...
> 
> Eric?
> 
> PS: My Perl is 5.8.0.

And the reference is from __gendep.  The problem is that it is
recursive.  I could not figure out how to not remove the prototype,
leave the function recursive, and not get that warning...

I guess we can remove the prototype so perl does not complain.

And it looks like I need to make certain undi_nii.c is checked in.

Eric

[Etherboot-developers] gendep() called too early?

[<ke...@us...>]

/usr/bin/perl ./genrules.pl i386 NIC  core/serial.c core/main.c
core/pci.c core/osloader.c core/nfs.c core/misc.c core/config.c
core/relocate.c core/heap.c drivers/disk/floppy.c core/nic.c core/disk.c
core/timer.c core/proto_slam.c core/proto_tftm.c
arch/i386/prefix/boot1a.s arch/i386/prefix/unhuf.S
arch/i386/prefix/unnrv2b.S arch/i386/firmware/pcbios/memsizes.c
arch/i386/firmware/pcbios/pcbios.S arch/i386/prefix/liloprefix.S
arch/i386/prefix/elfprefix.S arch/i386/prefix/comprefix.S
arch/i386/prefix/pxeprefix.S arch/i386/prefix/loader.S
arch/i386/core/start32.S arch/i386/core/start16.S arch/i386/core/pxe.c
arch/i386/core/pci_io.c arch/i386/core/i386_timer.c arch/i386/core/elf.c
> bin/Roms
main::__gendep() called too early to check prototype at ./genrules.pl line 48.
Driver file arch/ia64/drivers/net/undi_nii.c not found, skipping...

Eric?

PS: My Perl is 5.8.0.

Re: [Etherboot-developers] About Tulip

[<ke...@us...>]

>> I add some code in main.c and other .c,then I make bin32/an983.lzdsk,but it
>> report a error "ERROR:code size execeeds limits!",why and what can I do abo
>ut it
>
>Try it in etherboot-5.1.  For 5.0 the limits are very hard to up.
>In 5.1 the limit checking code is still there but the limits are higher,
>it is there simply to warn developers when there drivers are getting
>very big, and it will be difficult to fit them in a Rom chip.

It's actually not the ROM size limitation but the 48kB footprint
limitation of 5.0. However the size of the current image isn't near the
limit so he must have been doing things like adding large global
buffers.

> size bin32/tulip.tmp
  23237    2108   12968   38313    95a9 bin32/tulip.tmp

To "deng zhiguo", I think you're barking up the wrong tree anyway, if
the transfer finished but the image didn't execute, Etherboot worked and
it's more likely you didn't prepare your kernel and initrd properly.
Without more details we can't say. You don't even say what kind of
initrd it is. Is it even Linux, or are you trying to boot DOS?

Re: [Etherboot-developers] About Tulip

[<ebi...@ln...>]

"deng zhiguo" <dz...@ho...> writes:

> I add some code in main.c and other .c,then I make bin32/an983.lzdsk,but it
> report a error "ERROR:code size execeeds limits!",why and what can I do about it
> 
> ?

Try it in etherboot-5.1.  For 5.0 the limits are very hard to up.
In 5.1 the limit checking code is still there but the limits are higher,
it is there simply to warn developers when there drivers are getting
very big, and it will be difficult to fit them in a Rom chip.

Eric

[Etherboot-developers] About Tulip

[deng z. <dz...@ho...>]

I add some code in main.c and other .c,then I make bin32/an983.lzdsk,but it 
report a error "ERROR:code size execeeds limits!",why and what can I do 
about it ?
thanks very much!

_________________________________________________________________
与联机的朋友进行交流，请使用 MSN Messenger: http://messenger.msn.com/cn 

Re: [Etherboot-developers] About via-rhine.c

[Anselm M. H. <an...@ho...>]

Hello deng,

dz> I have a DFE-530TX NIC,I can download a image file from tftp server,but I 
dz> can not start it.The image file about 33Mb that was bulit with mknbi,why?

Let's ignore the fact that my Via-Rhine based card at least even did
not transfer 5 MB from Win2k via FTP, but look for where the problem
comes from.

How does it "not start"? Error messages?
You are sure that the transfer completed (perhaps last packet choked
or so, evtl see server log)?
The mknbi-Image really is OK? You know that you should use the newest
version of mknbi for the newest version of etherboot (BTW, which eb
version do you run)?

Best regards,
 Anselm                            mailto:an...@ho...

[Etherboot-developers] About via-rhine.c

[deng z. <dz...@ho...>]

I have a DFE-530TX NIC,I can download a image file from tftp server,but I 
can not start it.The image file about 33Mb that was bulit with mknbi,why?

_________________________________________________________________
免费下载 MSN Explorer:  http://explorer.msn.com/lccn/ 

[Etherboot-developers] [COMMIT] Working Itanium support

[<ebi...@ln...>]

I have finally gotten working Itanium support committed into the CVS
tree.

In addition I have modified the makefiles a little more, so that now
genrules.pl is passed $(SRCS) from which it generates dependencies
for all of the non-driver files.  Plus genrules.pl now correctly handles
recursive include files.  Previously it got the working directory incorrect.

The CVS version is bumped to 5.1.5-pre1 so that we don't get versions confused.

Eric

Re: [Etherboot-developers] Etherboot 5.1.4 (development) released

[<ebi...@ln...>]

ke...@us... writes:

> I've released Etherboot 5.1.4 at www.etherboot.org. It has support for
> the Itanium, major rearrangement of the files and some fixes, thanks to
> Eric Biederman. Please test it out and report experiences to this
> mailing list.
> 
> Remember that .lz* make targets are .z* targets now.

Small correction.  Most of the Itanium support is in but not quite
all of it.  There are a few parts I am still cleaning up, but I should
be finished merging into the CVS tree in the next couple of days...

Eric

[Etherboot-developers] Etherboot 5.1.4 (development) released

[<ke...@us...>]

I've released Etherboot 5.1.4 at www.etherboot.org. It has support for
the Itanium, major rearrangement of the files and some fixes, thanks to
Eric Biederman. Please test it out and report experiences to this
mailing list.

Remember that .lz* make targets are .z* targets now.