Syslinux - Dec 2004 - isolinux needing mkisofs

Due to limitations of equipment, I have need of using the isolinux boot 
loader with .iso files generated by roxio.

After trying (and of course failing) with an off the shelf version of 
isolinux, I grabbed the source for isolinux (syslinux 2.11 package) and 
started looking.

It took a few hours to orient myself, but I ended up doing the modification 
below to isolinux.asm.  I then had to write a small C program to take the 
.bin file and patch offset 0x14 with the proper checksum info.  With that I 
now have a isolinux.bin file suitable for roxio - and probably other 
software.

Is there any interest in this, or is the dependency on mkisofs there for 
another reason?

Thanks,

David



245c245
< bi_pvd:               dd 0x00000010                   ; LBA of primary 
volume descriptor      ;DEB
---
>bi_pvd:               dd 0xdeadbeef                   ; LBA of primary 
>volume descriptor
247c247
< bi_length:    dd size_of_code                 ; Length of boot file        
            ;DEB
---
>bi_length:    dd 0xdeadbeef                   ; Length of boot file
282a283,290
>               ; Set up boot file sizes
>               mov eax,[bi_length]
>               sub eax,SECTORSIZE-3
>               shr eax,2                       ; bytes->dwords
>               mov [ImageDwords],eax           ; boot file dwords
>               add eax,(2047 >> 2)
>               shr eax,9                       ; dwords->sectors
>               mov [ImageSectors],ax           ; boot file sectors
319,365d326
< ;DEB -- Patch to have code actually insert values for bi_file and 
bi_length
<               mov bx, trackbuf                ;DEB
<               mov eax, 17                     ;DEB Location of Boot Record 
Volume
<               call getonesec                  ;DEB Retrieve Boot Record 
Volume
<
<               mov eax, dword [trackbuf+47h]   ;DEB EAX is now address of 
Boot Catalog
<
< %ifdef DEBUG_MESSAGES
<               mov si,bcat_loc_msg             ;DEB Write sector address of 
boot cat
<               call writemsg                   ;DEB
<               call writehex8                  ;DEB
<               call crlf                       ;DEB
< %endif
<
<               mov bx, trackbuf                ;DEB
<               call getonesec                  ;DEB get the Boot Catalog
<
<               mov eax, dword [trackbuf+28h]   ;DEB put RBA sector address 
in EAX
<               mov dword [bi_file], eax        ;DEB Put RBA address into 
bi_file variable
<
< %ifdef DEBUG_MESSAGES
<               mov si,rba_loc_msg              ;DEB write out RBA sector 
address
<               call writemsg                   ;DEB
<               call writehex8                  ;DEB
<               call crlf                       ;DEB
< %endif
<
< %ifdef DEBUG_MESSAGES
<               mov si,imgsz_msg                ;DEB write out image size
<               call writemsg                   ;DEB
<               mov eax, dword [bi_length]      ;DEB
<               call writehex8                  ;DEB
<               call crlf                       ;DEB
< %endif                                                ;DEB
< ;DEB -- we now return you to your regularly scheduled programing :)
<
<
<               ; Set up boot file sizes
<               mov eax,[bi_length]
<               sub eax,SECTORSIZE-3
<               shr eax,2                       ; bytes->dwords
<               mov [ImageDwords],eax           ; boot file dwords
<               add eax,(2047 >> 2)
<               shr eax,9                       ; dwords->sectors
<               mov [ImageSectors],ax           ; boot file sectors
<
<
368d328
<
802,804d761
< bcat_loc_msg: db 'Boot Catalog found at ',0                        
;DEB
debug msg
< rba_loc_msg:  db 'RBA is ',0                                       
;DEB
debug msg
< imgsz_msg:    db 'Boot loader size in bytes ',0                    
;DEB
debug msg
1759,1760d1715
< size_of_code  equ ldlinux_end-bootsec         ;DEB Used to calculate 
bi_length

David Bond wrote:
> Due to limitations of equipment, I have need of using the isolinux boot 
> loader with .iso files generated by roxio.
> 
> After trying (and of course failing) with an off the shelf version of 
> isolinux, I grabbed the source for isolinux (syslinux 2.11 package) and 
> started looking.
> 
> It took a few hours to orient myself, but I ended up doing the 
> modification below to isolinux.asm.  I then had to write a small C 
> program to take the .bin file and patch offset 0x14 with the proper 
> checksum info.  With that I now have a isolinux.bin file suitable for 
> roxio - and probably other software.
> 
> Is there any interest in this, or is the dependency on mkisofs there for 
> another reason?
> 
Could you please send that as a unified diff?

The dependency on mkisofs is because the assumption that the BRV is at 
sector 17 isn't inherently valid.  The spec is that it's at sector 17 OF
THE LAST SESSION WRITTEN, which isn't the same thing.  This is basically 
a screwup in the El Torito specification, which is why the boot info 
block option to mkisofs exists to make up for it.

I don't know what you mean with "limitations of equipment";
mkisofs
should be able to run on pretty much anything, especially these days 
with MacOS being Unix and there being Cygwin for 'doze.

Or you can lobby the Roxio people for adding this extension, it's fully 
documented.

That being said, it wouldn't be a bad idea at all to add something like 
this as a fallback option; it's certainly easy to find if the info block 
hasn't already been patched, and it's better than failing completely. 
After all, it should work as long as it's not a multisession disk. 
Hence the request for a unified diff.

	-hpa

David Bond wrote:
> Here's a unified diff for a slightly different version. It checks to
see
> if the boot info table has been initialized by an external program.  If 
> not it then resorts to using sector 16 for the PVD and 17 for the BRV 
> and goes from there.  Otherwise it uses the values already present in 
> the boot info table.  This should be the best of all worlds per your 
> previous comments.
Agreed.  This got inlined into a piece of HTML, though, and came through 
butchered.  Could you send it as an attachment, perhaps (or at least 
turn off HTML in your MUA?)

	-hpa

Okay, after a bit more research I ended up implementing this in a 
slightly different way than you have.  In particular, there is at least 
an attempt on the part of the El Torito spec to address this by giving a 
pointer to the boot file (but not to the PVD which is the real problem.)

I have released this as 2.12-pre11; I would really appreciate it if you 
would try this out.

	-hpa