Syslinux - Oct 2010 - [syslinux:disklib] disklib: make CHS calculation match core/fs/diskio.c

On 10/13/2010 08:36 AM, syslinux-bot for Michal Soltys
wrote:
> Commit-ID:  9c8db7560e2dc83d1191bb2f90b4d4d0ae3d37d6
> Gitweb:    
http://syslinux.zytor.com/commit/9c8db7560e2dc83d1191bb2f90b4d4d0ae3d37d6
> Author:     Michal Soltys <soltys at ziu.info>
> AuthorDate: Wed, 13 Oct 2010 10:57:36 +0200
> Committer:  Michal Soltys <soltys at ziu.info>
> CommitDate: Wed, 13 Oct 2010 10:57:36 +0200
> 
> disklib: make CHS calculation match core/fs/diskio.c
> 
> Signed-off-by: Michal Soltys <soltys at ziu.info>
> 
I note that this code still:

a) duplicates the code between read and write, and
b) doesn't check for CHS overflow.

This is particularly critical for write, obviously, since we could
corrupt a disk.

	-hpa

On 10-10-13 23:30, H. Peter Anvin wrote:
> On 10/13/2010 08:36 AM, syslinux-bot for Michal Soltys wrote:
>>  Commit-ID:  9c8db7560e2dc83d1191bb2f90b4d4d0ae3d37d6
>>  Gitweb:    
http://syslinux.zytor.com/commit/9c8db7560e2dc83d1191bb2f90b4d4d0ae3d37d6
>>  Author:     Michal Soltys<soltys at ziu.info>
>>  AuthorDate: Wed, 13 Oct 2010 10:57:36 +0200
>>  Committer:  Michal Soltys<soltys at ziu.info>
>>  CommitDate: Wed, 13 Oct 2010 10:57:36 +0200
>>
>>  disklib: make CHS calculation match core/fs/diskio.c
>>
>>  Signed-off-by: Michal Soltys<soltys at ziu.info>
>>
>
> I note that this code still:
>
> a) duplicates the code between read and write, and
Oki, I'll shrink it.
> b) doesn't check for CHS overflow.
>
lbacnt set in disk_get_params() is always valid (if function is 
successful), and it's checked at the very begining of read/write 
functions (along with other sanity checks).

CHS check is no longer necessary, as CHS is used only if ebios == 0, and 
that implies that lbacnt is calculated from valid geometry (cbios == 1) 
or faked (1/1/1) in which lbacnt == 1.

I'll update the comments to make it more clear.

On Thu, Oct 14, 2010 at 9:08 AM, Shao Miller
<Shao.Miller at yrdsb.edu.on.ca> wrote:
> Somewhat unrelated: Some day in the far, far future, it might be nice to
> "re-image"[1] computers using INT 13h via one or more COMBOOT32
programs. ?-
> Shao
>
> [1] Apply a disk or filesystem snapshot/saved-state.
I did something similar for the resume usecase, i.e. I wrote a
resume.c32 module which resumes a computer from a hibernate image
(swsusp or TuxOnIce). This was used to boot an appliance from a known
state (the root filesystem being a read-only squashfs).

Blog entry here:
http://pub.mouraf.org/blog/2010/09/resume-c32-resume-linux-from-the-bootloader/
Code on github: http://github.com/pierre/syslinux-fastboot

-- 
Pierre-Alexandre Meyer
http://pub.mouraf.org/blog

Hello Shao, hpa,

I thought it was really ironic that you guys mentioned this.  I remember back
when I was imaging XP, I was trying to work out specifically what I'd want
to do to get gPXE/PXELINUX/COMBOOT to simplify our imaging process which at the
time was using a very clunky XP live CD that took forever to boot up...  I
ultimately ended up booting DOS from iSCSI and imaging via INT13 with Ghost 8,
which worked pretty well.

Still, while I was wrapping my head around the concepts I was trying to exploit,
I was browsing through all of these com32 modules in the SYSLINUX package then
hopped on IRC and said, "Hey guys, is there a COM32 module for imaging
disks?  Something like DD.c32?"

While several people (you guys included I think) remarked that it was an
interesting idea, I never found anything...  Hence the Ghost 8 use :-D

I do think it would be really nifty to have a DVD with ISOLINUX that boots to a
menu and images all via COM32...   no HAL issues or hardware compatibility
issues, no need to wait for a "thick" protected mode OS to load. 
Making modern operating systems easier to deploy by exploiting the ubiquity of
the "modern" BIOS implementations has a high geek-sexiness factor to
it ;)


Cheers,
Andrew Bobulsy



On Oct 16, 2010, at 3:00 PM, syslinux-request at zytor.com
wrote:
> 
> Message: 2
> Date: Fri, 15 Oct 2010 16:09:04 -0700
> From: "H. Peter Anvin" <hpa at zytor.com>
> To: For discussion of Syslinux and tftp-hpa <syslinux at zytor.com>
> Subject: Re: [syslinux] [syslinux:disklib] disklib: make CHS
> 	calculation	match core/fs/diskio.c
> Message-ID: <4CB8DF10.7030807 at zytor.com>
> Content-Type: text/plain; charset=ISO-8859-1
> 
> On 10/14/2010 09:08 AM, Shao Miller wrote:
>> 
>> Somewhat unrelated: Some day in the far, far future, it might be nice
to
>> "re-image"[1] computers using INT 13h via one or more
COMBOOT32
>> programs.  - Shao
>> 
>> [1] Apply a disk or filesystem snapshot/saved-state.
> 
> That's not even very hard at all, at least not for the whole-disk case.
> 
> 	-hpa