Step by step of my experience
Hi guys,
I've tried to boot Linux on my old clunky R5k this morning, and I thought I'd share the step by step with the list.
Basically, I've tried to boot off an EFS filesystem (d4), and here's how I went about it:
1. /d4 is my third scsi disk, with EFS (/dev/sdc under linux)
2. My irix kernel is in /unix on first disk (/dev/sda), which is XFS. Which is OK, since only the kernel is on XFS and the loader can handle it.
3. mkdir /d4/lsgi
4. Put ftp://ftp.linux.sgi.com/pub/test/vmlinux-970916-efs.gz in /d4/lsgi
5. Put ftp://ftp.linux.sgi.com:/pub/mips-linux/root-be-0.00.cpio.gz in /d4/lsgi
6. Extract root-be-0.00.cpio.gz in /d4:
cd /d4
gzip -dc lsgi/root-be-0.00.cpio.gz | cpio -icvd7. Install kernel on /
cd /d4/lsgi
gzip -d vmlinux-970916-efs.gz
cp vmlinux-970916-efs /vmlinux8. Reboot, go to maintenance mode, launch sash and type:
scsi(0)disk(1)rdisk(0)partition(0)/vmlinux root=/dev/sdc1
7. The kernel booted, said a great many (good) things and froze with the following messages:
VFS: mounted root (efs filesystem) readonly
EFS: inode 0x30301 has >EFS_MAX_EXTENTS (43)
EFS: <6> attempt to access beyond end of device 08:21 rw=0 want=1190802192, limit=4191574
Conclusion: Wait for Ralf to finish the EFS driver ! This is the exact bug he was talking about.
Also: I read a few mails about typing, under sash, the following:
boot /vmlinux root=/dev/sdc1
Well, it doesn't work.
First, the syntax for the sash boot command in my PROM is: boot -f <file> -n <args>
Second, when I did boot -f /vmlinux -n root=/dev/sdc1, it just loaded the kernel,
but didn't jump into it. It just told me: the entry address is <some hex number>.
When I did go <some hex number>, it ran linux, but tried to boot off tftp.
Obviously it ignored the args.
I'm just wondering: is there many different versions of sash out there ?
I'm off trying to boot off tftpd.
I'll keep the list posted.
Ciao,
- Mgix
Signature · 3 lines
________________________________________________________________________________ Emmanuel Mogenet <[email protected]> PGP Key on Request
Emmanuel,
Thanks for the report.
:
: EFS: <6> attempt to access beyond end of device
: 08:21 rw=0 want=1190802192, limit=4191574
:
:Conclusion: Wait for Ralf to finish the EFS driver ! :This is the exact bug he was talking about.
:
Actually, it is Mike Shaver, we talked about this one pretty recently, he is pretty busy but when he find some time to fix this it would make things much simpler.
Hope someone would be able to answer the sash booting question.
Signature · 1 line
Peace, Ariel
Thus spake Ariel Faigon:
Quoted 3 lines
Actually, it is Mike Shaver, we talked about this one pretty recently, he is pretty busy but when he find some time to fix this it would make things much simpler.
I've managed to procure a temporary Indy at work now, so I should be able to start hacking on EFS again early next week. I actually have better support already than that kernel would indicate, but I haven't committed all of the changes yet (I don't think).
I'll clean up my current support tomorrow or Friday, and then I'll check it in and coerce Miguel or Ralf into making a new test kernel available.
Mike
Signature · 5 lines
#> Mike Shaver ([email protected]) Ingenia Communications Corporation #> Chief System Architect and Herder of Bits #> #> "Yoda say, `Just slap a little public key crypto into it' does not #> a secure system make." -- Marcus J. Ranum ([email protected])
Quoted 4 lines
Actually, it is Mike Shaver, we talked about this one pretty recently, he is pretty busy but when he find some time to fix this it would make things much simpler.
Also the EFS code panics if its given an FS that isnt EFS it should just say "Nope". That confused me for a while
Anyway I've made a fair bit of progress by the time honoured technique of gluing other peoples code together in a new shape
I now have working SGI ext2fs tools that checkout v Linux ok. Ted has the patches.
I have a hacked up MacLinux installer which doesnt know tar/du/df/cp this time but does know how to unpack cpio files (only I broke /dev files and need to fix that)
With that installer and a bit of tweaking with mknod by hnd to fix the cpio device file bug I've got Linux/SGI up and running and all the base redhat bits on
Im being bitten by a couple of kernel bugs tho
1. vmalloc keeps failing on the unix GC (annoying noise)
2. SIGCLD isnt getting given to init for some reason
when scripts terminate (breaks redhat init scripting) [ note init isnt ignoring the signal... ]
Oh and does anyone happen to have a working ldconfig binary for Linux/SGI ?
Alan
Quoted 4 lines
Im being bitten by a couple of kernel bugs tho
1. vmalloc keeps failing on the unix GC (annoying noise)
The fix for this is commited in the cvs tree.
Quoted 3 lines
2. SIGCLD isnt getting given to init for some reason
when scripts terminate (breaks redhat init scripting) [ note init isnt ignoring the signal... ]
The latest SysV init works around this problem. Or at least fixes all the problems for me.
Cheers,
Miguel.
Quoted 2 lines
The latest SysV init works around this problem. Or at least fixes all the problems for me.
Ok I'll upgrade that when I build it as an rpm.
2.1.55 was fairly stable although rebuilding 3 rpms one of which is the howto's one (spends most of its time copying and compressing 200Mb of data) did finally get me a disk read from beyond the end of the disk.
A pile of rpms are on ftp.uk.linux.org:/pub/linux/incoming/SGI. I'll start playing collect the RedHat 4.9.1 rpms with other folks I guess now.
Alan