Re: Updated installer images
Hi Adrian,
On 09/08/2017 04:22 PM, John Paul Adrian Glaubitz wrote:
[...]
Will recreate the images in the evening. Sorry for that.
I yesterday tested the image from [1] (SHA256:
b07736dfa5bd167352d8f17f1d9bfc62fa9762f5496d0d134801b91f80bfb6b4) on an
Ultra 10. Sadly the boot loader installation still didn't work.
Unfortunately I also forgot to go back and check the available menu
items before finishing up the installation without boot loader
installation. I could later on install the boot loader manually via the
rescue mode, so a workaround is still possible with the current image.
[1]:
https://cdimage.debian.org/cdimage/ports/debian-9.0-sparc64-NETINST-1.iso
Attached are my notes. As the installer strangely didn't detect the
installed ATA HDD, although the kernel detected it and I could write to
it with dd, the failing boot loader installation could possibly also be
related to this issue.
I also had a look into the netinst installer images for i386 and amd64
for Debian 9.1.0. They only use one `pool` directory.
Why are two pool directories needed for sparc64? The `nobootloader` udeb
(all) is located below `pool`, the `silo-installer` udeb (sparc64) is
located below `pool-sparc64`. On the i386 installer ISO, e.g. the
`lilo-installer` udeb is located below `pool`, where there are both
`all` and `i386` packages available. So would it be possible to move or
just copy the `silo-installer` udeb to the pool directory structure for
testing?
Cheers and have a nice evening,
Frank
* the installer didn't detect the installed ATA HDD, although it's present in the kernel boot messages and I could write to it with busybox's dd in a shell. When I selected continue from the red screen, the disk could be used by the partitioner and I could continue with partitioning:
```
â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â?¤ [!!] Partition disks â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??
â?? â??
â?? This is an overview of your currently configured partitions and mount â??
â?? points. Select a partition to modify its settings (file system, mount â??
â?? point, etc.), a free space to create partitions, or a device to â??
â?? initialize its partition table. â??
â?? â??
â?? Guided partitioning - â??
â?? Configure software RAID â?? â??
â?? Configure the Logical Volume Manager â?? â??
â?? Configure ZFS â?? â??
â?? Configure encrypted volumes â?? â??
â?? Configure iSCSI volumes 0 â??
â?? â?? â??
â?? SCSI1 (0,0,0) (sda) - 40.0 GB ATA ST340810A â?? â??
â?? > 40.0 GB FREE SPACE . â??
â?? â??
â?? <Go Back> â??
â?? â??
â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??â??
<F1> for help; <Tab> moves; <Space> selects; <Enter> activates buttons
```
* Later on a kernel message displayed over the installer screen, but the installation continues. Checking in a shell yielded the following messages:
```
[ 3542.859672] mandb[2465]: segfault at 5c78 ip fffff801006238f4 (rpc fffff801006237f0) sp 000007feffc91ea1 error 30001 in libc-2.24.so[fffff8010056c000+15e000]
[ 3542.860190] BUG: Bad rss-counter state mm:fffff80000f1d740 idx:1 val:3
[ 3542.937756] BUG: non-zero nr_ptes on freeing mm: 1
```
I sometimes saw a rapidly increasing memory usage during mandb updates on Alpha machines with similar amounts of memory (512 MiB or less) which lead to the OOM killer starting to kill processes when no swap partition was used and heavy swapping when a swap partition was used. I'm unsure if this could be a similar issue.
* It again does not offer the boot loader installation, but this might be due to the same reason it did not detect any disks, though there were disks detected by the kernel.
Reply to: