* [parisc-linux] 712/60 boots fine
@ 2001-09-21 8:38 Andreas Härtel
2001-09-21 9:14 ` Richard Hirst
2001-09-21 16:03 ` Grant Grundler
0 siblings, 2 replies; 13+ messages in thread
From: Andreas Härtel @ 2001-09-21 8:38 UTC (permalink / raw)
To: Parisc-Linux (E-Mail)
Hi,
my 712 is running now. I've found my own mistake after many hours...
It is difficult to understand for a newbie, that PALO needs file system type
'F0' but this type is unknown by fdisk.
Or the 2. mistake, the partition where vmlinux is located must be lower or
equal 2 GByte.
Thanks
Andreas
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [parisc-linux] 712/60 boots fine
2001-09-21 8:38 [parisc-linux] 712/60 boots fine Andreas Härtel
@ 2001-09-21 9:14 ` Richard Hirst
2001-09-21 16:03 ` Grant Grundler
1 sibling, 0 replies; 13+ messages in thread
From: Richard Hirst @ 2001-09-21 9:14 UTC (permalink / raw)
To: Andreas Härtel; +Cc: Parisc-Linux (E-Mail)
On Fri, Sep 21, 2001 at 10:38:24AM +0200, Andreas Härtel wrote:
> Hi,
>
> my 712 is running now. I've found my own mistake after many hours...
>
> It is difficult to understand for a newbie, that PALO needs file system type
> 'F0' but this type is unknown by fdisk.
Partition type F0, not a file system type. Recent fdisk versions do
know F0 as "Linux/PA-RISC boot".
> Or the 2. mistake, the partition where vmlinux is located must be lower or
> equal 2 GByte.
Yes, that has caught me too in the past, I guess it should be
checked by palo or the installer.
Richard
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [parisc-linux] 712/60 boots fine
2001-09-21 8:38 [parisc-linux] 712/60 boots fine Andreas Härtel
2001-09-21 9:14 ` Richard Hirst
@ 2001-09-21 16:03 ` Grant Grundler
2001-09-21 16:42 ` Bdale Garbee
1 sibling, 1 reply; 13+ messages in thread
From: Grant Grundler @ 2001-09-21 16:03 UTC (permalink / raw)
To: Andreas Härtel; +Cc: Parisc-Linux (E-Mail)
=?iso-8859-1?Q?Andreas_H=E4rtel?= wrote:
> It is difficult to understand for a newbie, that PALO needs file system type
> 'F0' but this type is unknown by fdisk.
> Or the 2. mistake, the partition where vmlinux is located must be lower or
> equal 2 GByte.
Both issues are documented in the palo documentation.
See "man palo" (which points to /usr/share/doc/palo/).
The install ISO's also had/have a README.INSTALL that tells one about
both issues as well.
(Kudos to Paul Bame - he did an excellent job with both)
Is there a more obvious place this should be documented?
(ie where did you look?)
> Thanks
>
> Andreas
thanks,
grant
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [parisc-linux] 712/60 boots fine
2001-09-21 16:03 ` Grant Grundler
@ 2001-09-21 16:42 ` Bdale Garbee
2001-09-21 17:31 ` Grant Grundler
2001-09-21 21:04 ` Richard Hirst
0 siblings, 2 replies; 13+ messages in thread
From: Bdale Garbee @ 2001-09-21 16:42 UTC (permalink / raw)
To: parisc-linux
grundler@puffin.external.hp.com (Grant Grundler) writes:
> =?iso-8859-1?Q?Andreas_H=E4rtel?= wrote:
> > It is difficult to understand for a newbie, that PALO needs file system type
>> 'F0' but this type is unknown by fdisk.
>> Or the 2. mistake, the partition where vmlinux is located must be lower or
>> equal 2 GByte.
>
> Both issues are documented in the palo documentation.
Clearly, then, "just" having them documented is insufficient.
I noted when doing an install from the 0.9.2 CD myself the other day that we
paint a splash screen before disk partitioning explaining that there needs to
be a palo partition, but that splash screen doesn't mention the 2G issue.
Perhaps it could be updated with some additional text?
Also, since cfdisk is the default partitioning tool called by the Debian install
process, if the definition of partition type F0 isn't already part of the
list of partition types displayed by cfdisk when you pick the command to change
the partition type, perhaps it should get added? It might well be disconcerting
to a new user without much Debian or hppa experience to have to type in a
partition type that's not on the menu.
I think those two steps would improve the probability that folks are successful
installing.
Finally, it sounds as though the user-space palo executable is not checking all
of these constraints to warn the user about problems, though I haven't tried to
test that and so don't really know what it does or does not do. It certainly
seems prudent for palo to do so... and give the user some guidance through
appropriate warning/error text about what to do to fix the problem.
In summary, I think we ought to be thinking about how to help the user get this
right beyond "just" making sure the constraints are documented.
Bdale
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [parisc-linux] 712/60 boots fine
2001-09-21 16:42 ` Bdale Garbee
@ 2001-09-21 17:31 ` Grant Grundler
2001-09-21 19:17 ` thunder7
2001-09-21 21:04 ` Richard Hirst
1 sibling, 1 reply; 13+ messages in thread
From: Grant Grundler @ 2001-09-21 17:31 UTC (permalink / raw)
To: Bdale Garbee; +Cc: parisc-linux
Bdale Garbee wrote:
> Clearly, then, "just" having them documented is insufficient.
Agreed - that's why I asked Andreas where he looked.
> I noted when doing an install from the 0.9.2 CD myself the other day that we
> paint a splash screen before disk partitioning explaining that there needs to
> be a palo partition, but that splash screen doesn't mention the 2G issue.
> Perhaps it could be updated with some additional text?
Excellent idea. Who knows where the disk partitioning text is?
(and can someone sign up to "fix" it?)
Or perhaps Andreas can file a bug report? Where?
> In summary, I think we ought to be thinking about how to help the user get
> this right beyond "just" making sure the constraints are documented.
Certainly. But we had to start someplace. I'm happy Paul wrote
*any* documentation - an excellent set of documentation no less
and packaged the whole mess.
IMHO, new folks need to get involved for stuff like palo enhancements.
I'm sure Paul would welcome any patches people cook up.
Most of the original developers have moved on to other things and aren't
actively contributing in new features or enhancing existing documents/code.
This evolution seems natural since the types of issues outstanding today
are quite different than those from one or two years ago.
thanks,
grant
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [parisc-linux] 712/60 boots fine
2001-09-21 17:31 ` Grant Grundler
@ 2001-09-21 19:17 ` thunder7
2001-09-25 1:50 ` Grant Grundler
0 siblings, 1 reply; 13+ messages in thread
From: thunder7 @ 2001-09-21 19:17 UTC (permalink / raw)
To: parisc-linux
On Fri, Sep 21, 2001 at 11:31:51AM -0600, Grant Grundler wrote:
> Certainly. But we had to start someplace. I'm happy Paul wrote
> *any* documentation - an excellent set of documentation no less
> and packaged the whole mess.
>
> IMHO, new folks need to get involved for stuff like palo enhancements.
> I'm sure Paul would welcome any patches people cook up.
>
I guess that means me :-)
This is my current palo output:
jurriaan@pa8200:~$ sudo /sbin/palo
palo version 0.94 bame@c3k Sat Sep 8 12:33:24 MDT 2001
ELF32 executable
Partition Start(MB) End(MB) Id Type
1 1 31 f0 Palo
2 32 3846 83 ext2
3 3847 7661 83 ext2
4 7662 8678 82 swap
ipl: addr 16384 size 28672 entry 0x0
ko 0x0 ksz 0 k64o 0x0 k64sz 0 rdo 0 rdsz 0
<2/boot/vmlinux-2.4.9-pa24 root=/dev/sda2 HOME=/>
ipl: addr 16384 size 28672 entry 0x0
ko 0x44000 ksz 2853588 k64o 0x0 k64sz 0 rdo 0 rdsz 0
<2/boot/vmlinux-2.4.9-pa24 root=/dev/sda2 HOME=/>
Now if I understand correctly, palo must
- warn if there is no f0 partition
- warn if there is no ext2 partition in the first 2 Gb
The first is not that difficult, the relevant patch is here:
===================================================
diff -Br -b -U 3 -N palo/lib/diskpart.c palo-new/lib/diskpart.c
--- palo/lib/diskpart.c Fri Jun 15 12:05:10 2001
+++ palo-new/lib/diskpart.c Fri Sep 21 22:37:54 2001
@@ -97,6 +97,7 @@
{
int i;
const int mbshift = 20 - 9;
+ int f0_found = 0;
printf("Partition Start(MB) End(MB) Id Type\n\r");
for (i = 0; i < maxparts; i++)
@@ -110,6 +111,16 @@
mptab[i].id == 0x83 ? "ext2" :
(mptab[i].id == 0x82 ? "swap" :
(mptab[i].id == 0xf0 ? "Palo" : "Unknown")));
+ if ( (mptab[i].id == 0xf0) &&
+ (1 + ((mptab[i].start + mptab[i].length) >> mbshift) < 2048) )
+ f0_found = 1;
+ }
+ if (f0_found == 0)
+ {
+ printf("WARNING: You have no partition with type f0 within the first 2 Gb of the disk.\n");
+ printf("palo needs this partition to store the bootloader and optional (recovery)\n");
+ printf("kernel/ramdisk. It's size should be 4 (minimal) to 16 (safe) Mb.\n");
+ printf("Read /usr/share/doc/palo/README for more information.\n");
}
}
===================================================
The second problem should probably be something like this:
===================================================
diff -Br -b -U 3 -N palo/palo/palo.c palo-new/palo/palo.c
--- palo/palo/palo.c Thu Jun 21 12:05:38 2001
+++ palo-new/palo/palo.c Fri Sep 21 23:14:48 2001
@@ -784,6 +792,20 @@
}
if (f0 == MAXPARTS)
error(11);
+ if (strlen(commandline) > 1)
+ {
+ int bootpart=atoi(commandline);
+ if ((bootpart > 0) && (bootpart < MAXPARTS))
+ {
+ /* 2^11 = 1048576 / 512 */
+ if (1+((ptab[bootpart].start + ptab[bootpart].length) >> 11) >= 2048 )
+ {
+ printf("Warning: your kernel-image ('%s') is on a partition that isn't entirely below 2 Gb.\n",commandline);
+ printf("Palo can't read files beyond 2 Gb on older machines. Read /usr/share/doc/palo/README for more informa
tion.\n");
+ }
+ }
+ }
+
print_ptab_pretty(ptab, MAXPARTS);
if (0) printf("F0 partition start sector %d length %d\n",
===================================================
I'd appreciate pointers to improve this.
If this works out, perhaps adding reiserfs support to palo is next?
It works in yaboot (PPC boot-loader), after all.
Good luck,
Jurriaan
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
test.out: unmodified: line 1
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
~
test.out: unmodified: line 1
--
Your face is just a mask you wear,
but masks are hidden faces
The Oysterband - The road to Santiago
GNU/Linux 2.4.9-ac10 SMP/ReiserFS 2x1402 bogomips load av: 0.00 0.00 0.00
^ permalink raw reply [flat|nested] 13+ messages in thread* Re: [parisc-linux] 712/60 boots fine
2001-09-21 19:17 ` thunder7
@ 2001-09-25 1:50 ` Grant Grundler
0 siblings, 0 replies; 13+ messages in thread
From: Grant Grundler @ 2001-09-25 1:50 UTC (permalink / raw)
To: thunder7; +Cc: parisc-linux
thunder7@xs4all.nl wrote:
> I guess that means me :-)
Uhm...yeah. :^)
> This is my current palo output:
>
> jurriaan@pa8200:~$ sudo /sbin/palo
> palo version 0.94 bame@c3k Sat Sep 8 12:33:24 MDT 2001
> ELF32 executable
> Partition Start(MB) End(MB) Id Type
> 1 1 31 f0 Palo
> 2 32 3846 83 ext2
> 3 3847 7661 83 ext2
> 4 7662 8678 82 swap
> ipl: addr 16384 size 28672 entry 0x0
> ko 0x0 ksz 0 k64o 0x0 k64sz 0 rdo 0 rdsz 0
> <2/boot/vmlinux-2.4.9-pa24 root=/dev/sda2 HOME=/>
> ipl: addr 16384 size 28672 entry 0x0
> ko 0x44000 ksz 2853588 k64o 0x0 k64sz 0 rdo 0 rdsz 0
> <2/boot/vmlinux-2.4.9-pa24 root=/dev/sda2 HOME=/>
>
> Now if I understand correctly, palo must
>
> - warn if there is no f0 partition
> - warn if there is no ext2 partition in the first 2 Gb
warn if the partition with vmlinux (eg /boot/vmlinux) exceeds 2GB.
The problem is palo uses IODC to load data.
IODC can't access beyond 2GB due to signed numbers/offset issues.
Note that your partition 2 violates that rule.
grant
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [parisc-linux] 712/60 boots fine
2001-09-21 16:42 ` Bdale Garbee
2001-09-21 17:31 ` Grant Grundler
@ 2001-09-21 21:04 ` Richard Hirst
1 sibling, 0 replies; 13+ messages in thread
From: Richard Hirst @ 2001-09-21 21:04 UTC (permalink / raw)
To: Bdale Garbee; +Cc: parisc-linux
On Fri, Sep 21, 2001 at 10:42:51AM -0600, Bdale Garbee wrote:
> grundler@puffin.external.hp.com (Grant Grundler) writes:
>
> > =?iso-8859-1?Q?Andreas_H=E4rtel?= wrote:
> > > It is difficult to understand for a newbie, that PALO needs file system type
> >> 'F0' but this type is unknown by fdisk.
> >> Or the 2. mistake, the partition where vmlinux is located must be lower or
> >> equal 2 GByte.
> >
> > Both issues are documented in the palo documentation.
>
> Clearly, then, "just" having them documented is insufficient.
>
> I noted when doing an install from the 0.9.2 CD myself the other day that we
> paint a splash screen before disk partitioning explaining that there needs to
> be a palo partition, but that splash screen doesn't mention the 2G issue.
> Perhaps it could be updated with some additional text?
Done. Now says:
| PALO, the PArisc LOader, requires a partition of type 'f0' on the disk #
| it is loaded from. Therefore, if this disk will contain the boot #
| loader you must create a partition of type 'f0' of at least 16MB. #
| #
| A further important restriction is that the partition you load your #
| kernel from must reside within the first 2GB of your disk. The kernel #
| lives in /boot, so you can either keep your root file system within the #
| first 2GB, or you can create a small partition and mount it as /boot. #
> Also, since cfdisk is the default partitioning tool called by the Debian install
> process, if the definition of partition type F0 isn't already part of the
> list of partition types displayed by cfdisk when you pick the command to change
> the partition type, perhaps it should get added? It might well be disconcerting
cfdisk 2.11h
Disk Drive: /dev/sda
Size: 9100044288 bytes
Heads: 64 Sectors per Track: 32 Cylinders: 8678
Name Flags Part Type FS Type [Label] Size (MB)
------------------------------------------------------------------------------
sda1 Primary Linux swap 135.27
sda2 Primary Linux/PA-RISC boot 17.83
sda3 Primary Linux ext2 1999.64
Pri/Log Free Space 6946.82
So F0 is recognised as "Linux/PA-RISC boot" now, at least.
Richard
^ permalink raw reply [flat|nested] 13+ messages in thread
* [parisc-linux] 715/50
@ 2002-01-13 20:34 geezer
2002-01-13 22:07 ` [parisc-linux] Re: sid vs woody (was 715/50) Grant Grundler
0 siblings, 1 reply; 13+ messages in thread
From: geezer @ 2002-01-13 20:34 UTC (permalink / raw)
To: parisc-linux, debian-hppa
[-- Attachment #1: Type: text/plain, Size: 305 bytes --]
So the palinux-0.9-README file says "YOU MUST INSTALL 'sid'", blah, blah, blah...
I get "failed getting release file file:/instmnt/dists/sid/Release"
The woody one installs OK, but I am not sure how "crippled" I may be with it.
Any ideas? What more can I tell you to help with the diagnosis?
Geezer
[-- Attachment #2: Type: text/html, Size: 811 bytes --]
^ permalink raw reply [flat|nested] 13+ messages in thread
* [parisc-linux] Re: sid vs woody (was 715/50)
2002-01-13 20:34 [parisc-linux] 715/50 geezer
@ 2002-01-13 22:07 ` Grant Grundler
2002-01-14 10:47 ` Andreas Deresch
0 siblings, 1 reply; 13+ messages in thread
From: Grant Grundler @ 2002-01-13 22:07 UTC (permalink / raw)
To: geezer; +Cc: parisc-linux, debian-hppa
"geezer" wrote:
> So the palinux-0.9-README file says "YOU MUST INSTALL 'sid'", blah, =
> blah, blah...
Uhm...which release? URL or from an ISO?
> I get "failed getting release file file:/instmnt/dists/sid/Release"
> The woody one installs OK, but I am not sure how "crippled" I may be =
> with it.
> Any ideas? What more can I tell you to help with the diagnosis?
For 0.9.3, woody is correct. You won't be very crippled at all.
With "sid", plan on hppa being broken occasionally. I expect "woody"
to be more stable - IMHO better for folks who just want to use debian.
Newer kernel (compared to 0.9.3 ISO) from ftp.parisc-linux.org
will work for either release.
For the adventerous, edit the /etc/apt/sources.list to point
to "sid" like this:
deb http://http.us.debian.org/debian sid main contrib non-free
deb http://non-us.debian.org/debian-non-US sid/non-US main contrib non-free
Then do "apt-get update" and "apt-get dist-upgrade".
grant
^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: [parisc-linux] Re: sid vs woody (was 715/50)
2002-01-13 22:07 ` [parisc-linux] Re: sid vs woody (was 715/50) Grant Grundler
@ 2002-01-14 10:47 ` Andreas Deresch
2002-01-14 17:42 ` Bdale Garbee
0 siblings, 1 reply; 13+ messages in thread
From: Andreas Deresch @ 2002-01-14 10:47 UTC (permalink / raw)
To: Grant Grundler; +Cc: parisc-linux
> For the adventerous, edit the /etc/apt/sources.list to point
~~~~~~~~~~~
> to "sid" like this:
>
> deb http://http.us.debian.org/debian sid main contrib non-free
> deb http://non-us.debian.org/debian-non-US sid/non-US main contrib non-free
Shouldn't one use stable, testing and unstable, which are of course at the
moment linked to potato, woody and sid respectively? Otherwise you will stay
with sid even when it is declared stable (or unavailable) - and you don't
want that, do you? ;-)
ad
^ permalink raw reply [flat|nested] 13+ messages in thread* Re: [parisc-linux] Re: sid vs woody (was 715/50)
2002-01-14 10:47 ` Andreas Deresch
@ 2002-01-14 17:42 ` Bdale Garbee
2002-01-14 18:41 ` Grant Grundler
0 siblings, 1 reply; 13+ messages in thread
From: Bdale Garbee @ 2002-01-14 17:42 UTC (permalink / raw)
To: parisc-linux
aderesch@fs.tum.de (Andreas Deresch) writes:
> > deb http://http.us.debian.org/debian sid main contrib non-free
> > deb http://non-us.debian.org/debian-non-US sid/non-US main contrib non-free
> Shouldn't one use stable, testing and unstable, which are of course at the
> moment linked to potato, woody and sid respectively? Otherwise you will stay
> with sid even when it is declared stable (or unavailable) - and you don't
> want that, do you? ;-)
"It depends."
First of all, sid won't ever be stable. Sid was the kid in the Toy Story
movie who was sort of scary and should never be let out of the house... (well,
actually he's the character I most identified with in the movie, but we won't
go there... :-) and it's the *permanent* codename for Debian unstable. In
our new pool-based scheme, new names get declared for the 'testing' release
and become stable with time. So, 'woody' is now the same as 'testing' and
will be the same as 'stable' when it is released... at which point a new name
will be assigned for the new testing release, and so on.
Sometimes it makes sense to use the release tokens. You may want to always
run 'stable' on a machine to get the latest stable release, for example. In
other cases, it makes much more sense to track a particular named release and
have conscious control over when you hop from one release to another. If you
want to track unstable, either 'sid' or 'unstable' is fine. If you want to
start with woody now and stay with it when it goes stable, 'woody' might be
better than 'testing' to use in the sources.list file.
Hope that helps. For what it's worth, I always use the code names and exert
explicit control over when I hop from one release to another... but I've helped
configure production servers with files that say 'stable' and also include
entries for the security.debian.org updates (crucial if tracking stable!).
Bdale
^ permalink raw reply [flat|nested] 13+ messages in thread
end of thread, other threads:[~2002-01-14 18:41 UTC | newest]
Thread overview: 13+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2001-09-21 8:38 [parisc-linux] 712/60 boots fine Andreas Härtel
2001-09-21 9:14 ` Richard Hirst
2001-09-21 16:03 ` Grant Grundler
2001-09-21 16:42 ` Bdale Garbee
2001-09-21 17:31 ` Grant Grundler
2001-09-21 19:17 ` thunder7
2001-09-25 1:50 ` Grant Grundler
2001-09-21 21:04 ` Richard Hirst
-- strict thread matches above, loose matches on Subject: below --
2002-01-13 20:34 [parisc-linux] 715/50 geezer
2002-01-13 22:07 ` [parisc-linux] Re: sid vs woody (was 715/50) Grant Grundler
2002-01-14 10:47 ` Andreas Deresch
2002-01-14 17:42 ` Bdale Garbee
2002-01-14 18:41 ` Grant Grundler
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox