See URL above for complete info. In summary, the install instructions call for partitioning the /boot (or /) partition starting at block 1, and the Sun "whole disk" partition at block 0. When SILO is run, it installs the boot info at block 0. If PROM is set to boot from partition 1 (e.g., /dev/hda1 for IDE), which is usually the case, it will search for the boot block at block 1 (the beginning of partition 1) and not block 0, resulting in the error message "The file just loaded does not appear to be executable." The only partition containing block 0 is the "whole disk" partition, therefore, PROM must be told to boot from that (by default, according to the SPARC install guide, partition 3) by issuing "boot disk:c" or something analogous at the OK prompt. This can be modified to be the default behavior by issuing "setenv boot-dev disk:c [floppy] [cdrom] [net]" at the OK prompt. Recommend adding this info to Section 29 of the SPARC installation guide (http://www.gentoo.org/doc/en/gentoo-sparc-install.xml#doc_chap29). Reproducible: Always Steps to Reproduce: 1. Follow install instructions to the letter. Actual Results: PROM reports: "The file just loaded does not appear to be executable" Expected Results: SILO should boot. This is using Open Boot PROM (OBP) 3.31.0 on an Ultra 5 with IDE disks.
Created attachment 21196 [details, diff] Patch to gentoo-sparc-install.xml This should take care of it. If a docdev and/or sparc dev can review?
Created attachment 21198 [details, diff] Patch to hb-install-finalise.xml
Might want to follow that "setenv" command with either a "boot" or "reset" to complete the boot. Also, I'm not sure what the appropriate values would be for a SCSI system (or for that matter a SPARC32 system). In my Open Boot PROM (v. 3.31.0 on Ultra 5) the NVRAM defaults assigned correct aliases to "disk" and "disk:a" through "disk:d" (maybe through "disk:h" but I'm hesitant to check now because I'm compiling). I'm not sure if this is standard for other configurations.
Created attachment 21205 [details, diff] Patch to gentoo-sparc-install.xml
Created attachment 21206 [details, diff] Patch to hb-install-finalise.xml
Can a sparc dev (someone with doc experience -- Joe?) review these?
Has this been verified? The "by-the-book" install should work. I have run into having to specify "boot disk:a" for example, but that's usually a silo.conf or SILO config error. Can anybody else replicate this? I would be concerned with adding a doc note about changing a PROM variable via set-env, since that is saved in prom and may cause issues with any subsequent disk installs on that disk.
I can totally agree on your points; does this mean that the forum (and in se the bugreport here) is invalid? I'm ofcourse totally in the dark here, I cannot reproduce this...
I've never had to do this. I suspect that it's only an issue if the disk is not set up correctly. Could someone who's needed to do this please post `fdisk -l /dev/discs/disc0/disc`?
Created attachment 21291 [details] Output of fdisk -l /dev/discs/disc0/disc I know you were looking for someone else's, but FWIW, here's mine for verification. Also, I was incorrect about the default aliases above. They are not set up for partitions, but rather for disks, so I have disk, disk1, disk2, disk3 -- pointing to the IDE disks in order. The colon notation is apparently native to PROM for designating partitions, i.e. there is no disk:c alias. Regarding setenv, you can just suggest "boot disk:c" and leave it as an exercise to the reader to figure out how to make PROM do that automagically.
Perhaps it is better to suggest it as a solution to a possible problem rather than make it a difinitive step, and to use boot disk:c instead of changing the NVRAM variable. It is quite possible that this only manifests itself on IDE systems or some other small subset, and you could confuse others (and perhaps break their boots) by suggesting the procedure as mandatory before it's determined that the by-the-book install doesn't work.
I'm not in favor of using the docs to fix a problem that isn't documentation related :) So unless I am confident that this situation is not a user error because (s)he didn't read the documents correctly, I wont put this in the installation instructions.
This is something I see come up occasionally, but I have not found out what causes you to have to this. None of the machines I've used (scsi and ide based) have needed this. If we could narrow down why this is happening, we could make a note about that in the docs. On a side note, you can make an nvram alias to point to a particular disk and partiion. An example of this is; Currently, an Ultra 5 I use sees the disk1 alias as /pci/pci@1,1/ide/disk@0,0:a Now, if we need to boot off of partition c (or partition 3 as linux sees it), then the following command would create a new alias called gentoo that points to partition c; ok nvalias gentoo /pci/pci@1,1/ide/disk@0,0:c Then you can set your boot-device to the alias gentoo ok setenv boot-device gentoo Then reset the box to make sure the changes have taken effect ok reset
All nice and well, but it is perhaps better to write up a SPARC FAQ? Or SPARC Tips & Tricks? This is something all architectures can benefit from imho. I'm sure the SPARC team has a nice mailbox filled with Q&A that they could place on some special document to refer to :)
Yeah, Blademan and I had started one, need to ping him on the status.
I'm going to mark this as INVALID. As long as I don't hear from the sparc team, I can't act. Sparc-dudes, if the SPARC Faq/Tips 'n Tricks is somewhere, just ping me :)
Forgot to select the right option :/
