Bug 31012

From https://bugs.mageia.org/show_bug.cgi?id=28409#c10
"For the record, the 32-bit Linux kernel does not support booting from a
64-bit UEFI BIOS, which is why the 32-bit ISOs don't include a 64-bit EFI
bootloader."

CC: (none) => davidwhodgins

Oops. Forgot to close.

Resolution: (none) => INVALID
Status: NEW => RESOLVED

yes,

issue is that this unit has a 32bit UEFI.
Mga8 32bit isos boot and install on this hardware, it was specifically acquired for these tests.

Ah. Sorry, reopening. Assigning to kernel and drivers team.

Status: RESOLVED => REOPENED
Resolution: INVALID => (none)
Assignee: bugsquad => kernel

@Ben, after the error occurs, if you use Ctrl-Alt-F4 to switch to the screen that displays the kernel messages, do you see an error message like this:

traps: runinstall2[447] trap invalid opcode ip:b5f361d1 sp:bfa9f940 error:0 in libffi.so.0[b5f33000+6000]

For me, on an Asus X205TA, this is occurring almost immediately after the installer switches to stage 2 and before the GUI starts. I don't see any error mesages about no disk drives found or no module list found, but that may just be a timing difference.

CC: (none) => mageia

(In reply to Martin Whitaker from comment #5)

> For me, on an Asus X205TA, this is occurring almost immediately after the
> installer switches to stage 2 and before the GUI starts.

this is where it occurs for me, just before the GUI appears.

In an effort to discount the usb stick, I put the .iso on the eMMC device and attempted to use that as my install source (harddisk), but it couldn't find it: 
*Error:no disk drive is found*

am away from that unit for several hours, will let you know about the ctrl-alt-f4

(In reply to Martin Whitaker from comment #5)
> @Ben, after the error occurs, if you use Ctrl-Alt-F4 to switch to the screen
> that displays the kernel messages, do you see an error message like this:
> 
> traps: runinstall2[447] trap invalid opcode ip:b5f361d1 sp:bfa9f940 error:0
> in libffi.so.0[b5f33000+6000]
> 

essentially very similar:
traps: runinstall2[460] trap invalid opcode ip:b5ed71d1 sp:bfec5760 error:0
in libffi.so.8[b5ed4000+6000]

hmmm.

also happens on an elderly Compact Presario C700, not UEFI capable by any means!

traps: runinstall2[415] trap invalid opcode ip:b5f111d1 sp:bfaf9990 error:0
in libffi.so.8[b5f0e000+6000]

cpu : celeron m cpu 530 @1.73GHz

SSD : i-Flashdisk_K8_64GB

transferred the USB stick to my Toshiba and it booted to stage 2, implying it is not the stick.

tried from a real DVD, same.

and a HP Compaq 8510w

The Celeron M 530 seems reasonably recent (2007, 64 bit, SSE3).
The Asus X205TA has an even more recent CPU (2014, SSE4.2).
I really wonder the invalid opcode, I'll try to check the binary later today or tomorrow.

Yes, it looks like the code is jumping to the wrong address rather than it being a real unsupported instruction. Oddly this bug doesn't show up in VirtualBox, but does in QEMU on the same host machine.

Updating drakx-installer-stage2 to adapt to the new version of WebKit2 appears to fix the bug. Just running a full install on the X205TA to be sure (I'd forgotten how slow it is :-( ).

ah yes. Virtualbox.

one of *those* quirks!

try a Linx-7 for speedy (not)
you need to start with a fully charged battery, as you cannot have external power and install at the same time.

this processor is also affected:

Intel(R) Core(TM) i3 CPU       M 350  @ 2.27GHz
(Toshiba Satellite L650)

The fix will be included in the next ISO build.

Status comment: (none) => fixed in drakx-installer-stage2-18.51
Source RPM: (none) => drakx-installer-stage2

thanks Martin.
will look forward to them.

Just tested the 4 CPUs that initial .iso failed to install on.

for all 4, UEFi and legacy, installer runs to at least the  language selection (stage-2)

marking as resolved

Resolution: (none) => FIXED
Status: REOPENED => RESOLVED