Hi,
after hours of kernel compilations I think I've found the cause for the
hang with the message "entry point at 0x1001000".
It seems that the kernel is to large to boot.
Kernels (/bsd) larger than about 18 MB won't boot on this machine.
The last successful kernel compilation and boot was with a cvs checkout
from 2020-06-27. All kernel builds after this date don't boot.
First I thought it was the efifb change
http://cvsweb.openbsd.org/cgi-bin/cvsweb/src/sys/arch/amd64/amd64/efifb.c?f=h
After that I thought it was the umstc and acpihid device driver.
But that not true :)
I can run a current kernel when I reduce the kernel size by disabling
the amdgpu and radeondrm devices in the GENERIC configuration file.
I like to stay on -current by running regularly sysupgrade. But I don't
want to recompile the kernel to reduce the kernel size.
Also disabling the amdgpu and radeondrm devices via boot_config doesn't
work.
Do you have an idea if its possible to specify the memory areas which
the loader should choose to load the kernel, e.g. boot> machine memory
=<some size>M ?
My machine memory looks like this:
Region 0: type 1 at 0x0 for 352KB
Region 1: type 2 at 0x58000 for 4KB
Region 2: type 1 at 0x59000 for 180KB
Region 3: type 2 at 0x86000 for 488KB
Region 4: type 1 at 0x100000 for 261120KB
...
The kernel doesn't boot/hang at "entry point 0x100100" with following (I
think memory mappings):
booting sr0a:bsd: 14447944+3175440+340000+0+...
The kernel boots successfully
Default 6.7 MP kernel
booting sr0a:bsd: 12952904+2753544+331808+0+...
and also my custom kernel without radeondrm und amdgpu
booting sr0a:bsd: 10409288+2253840+306784+0+...
Thanks and best regards,
Sven
On 6/24/20 6:02 PM, Sven Wolf wrote:
> Hi,
>
> I did some more tests/new installations on this machine.
>
> A clean reinstallation of 6.7 release boots without any problems with
> the loaders 3.50 (release 6.7) and also 3.52 (snapshot 2020-06-23).
> Kernel 6.7 GENERIC.MP#182 build time 2020-05-07
>
> Also after an syspatch the kernel 6.7 GENERIC.MP#2 build time 2020-06-04
> boots without any problems with the loaders 3.50 and 3.52 (snapshot
> 2020-06-23).
>
> So in my opinion the problem is the kernel. I think between 2020-06-04
> and 2020-06-23 the where some changes which prevents that the 6.7 and
> 6.7/current loader boots the kernel on this and maybe on some other
> Lenovo (e.g. X1 Gen7) machines.
>
> The loader always show "entry point at 0x1001000" for some milliseconds,
> also when the kernel 6.7 gets loaded successfully.
>
> Maybe we'll find the changes for the kernel, which are responsible for
> this error. An another point for the kernel issue is in my opinion that
> the current RAMDISK kernel (from the latest snapshot) boots also without
> any problems.
>
> I'll do some more testing with a current kernel.
>
> Best regards,
> Sven
>
> On 6/23/20 10:20 PM, Matt Kunkel wrote:
>> Here is the offending patch. -current boots fine with it removed:
>>
>> http://cvsweb.openbsd.org/cgi-bin/cvsweb/src/sys/arch/amd64/stand/efiboot/exec_i386.c.diff?r1=1.2&r2=1.3&f=h
>>
>>
>> Appears bootx64.efi is carrying some board specific workaround for
>> HP / Computrace that breaks many others, including tianocore. Guess
>> I'll submit a patch to reverse it since I don't have an Elitebook to
>> test with?
>>
>> As a workaround, mount the efi partition and copy in boot64.efi from
>> 6.6.
>>
>> -Matt Kunkel
>>
>> June 23, 2020 2:16 PM, "Sven Wolf" <mailinglist@fusion-zone.net> wrote:
>>
>>> Hi,
>>>
>>> also after the new installation of the current snapshot the system
>>> stops with "entry point at
>>> 0x1001000".
>>> It's interesting, that a installation via bsd.rd is possible. But
>>> after that the system doesn't
>>> boot via bsd.mp/bsd.sp.
>>>
>>> Best regards,
>>> Sven
>>>
>>> On 6/21/20 8:55 PM, Sven Wolf wrote:
>>>
>>>> Hi,
>>>> the update of the loader didn't help.
>>>> I've updated the bootx64.efi from 3.48 to 3.52. But the current
>>>> kernel > doesn't load. I'll try a
>>>> re-installation.
>>>> Maybe @Otto can explain why the start of bsd.rd is possible and the
>>>> > start of bsd.sp/bsd.mp is not
>>>> possible. Maybe I can build a custom kernel.
>>>> Best regards,
>>>> Sven
>>>>> On 6/21/20 8:33 PM, Sven Wolf wrote:
>>>>> Hi,
>>>>>
>>>>> I found the same issue in a thread some weeks ago.
>>>>> https://marc.info/?l=openbsd-misc&m=159039904132502&w=2
>>>>>
>>>>> I'll test an reinstall/older loader. Boot from mbr isn't an option :(
>>>>>
>>>>> Best regards,
>>>>> Sven
>>>>>
>>>>> On 6/21/20 8:20 PM, Sven Wolf wrote:
>>>>
>>>> Hi,
>>>>
>>>> I've upgraded my Lenovo V130 from snapshot 6.6 (April 2020) to the
>>>> >>> snapshot from 2020-06-20.
>>>> The boot via boot.rd is always possible.
>>>> But when I load bsd.sp or bsd.rd the boot process stops with the >>>
>>>> error "entry point 0x1001000".
>>>> Do you have an idea how I can fix this >>> error?
>>>> In the past I did't have any problem with openbsd on this machine.
>>>>
>>>> I'll try tomorrow the next snapshot.
>>>>
>>>> Thanks and best regards,
>>>> Sven
>>>>
>>>>>
>>
>
No comments:
Post a Comment