Re: Bogus low-memory mode
On Tue, May 28, 2013 at 5:22 PM, Samuel Thibault <sthibault@debian.org> wrote:
> Donald Allen, le Tue 28 May 2013 17:17:34 -0400, a écrit :
>> I am attempting to install the Debian Gnu/Hurd release on circa 2005
>> Thinkpad G41, 2 GB of memory and a 60 GB IDE disk. Early in the the
>> process, I got a dire warning about insufficient memory, telling me
>> that something like 393 MB of memory are required and that if I
>> proceed, bad things are likely to happen.
>
> Uh?
I had the same reaction.
>
> Could you switch to console 2, run vmstat-hurd and tell us the figures?
> Could you also check at the beginning of /var/log/dmesg the "physical
> memory from 0x0 to ..." figures?
It also occurred to me to have a look at that upper memory limit that
went by early as the install kernel was booting. It's 0xf6f0000 =
258932736. No wonder the installer is complaining.
I also tried the install on an even older Thinkpad 600x I have with
512 MB of memory, and that worked correctly. The upper limit was
0x23fd0000 = 603783168.
No /var/log/dmesg file, by the way. There is a /var/log/syslog and
that's where I got the upper limit.
To get you your vmstat-hurd output from the G41, since I didn't
(couldn't) complete the install, I booted the install DVD and aborted
the install to a shell at the first opportunity (no option to get to a
shell on the initial screen?).
pagesize : 4096
size: 240M
active: 8204K
inactive: 89.8
wired: 21.2M
zero filled: 276M
reactivated: 0
pageins: 8440K
pageouts: 3564K
page faults: 544659
cow faults: 74487
memobj hit ratio: 99%
cached momobjs: 99
cache: 516K
swap size: 0
swap free: 0
>> I realize that this project is very resource-constrained, but a bug
>> like this on very standard hardware make me wonder how much testing
>> was done prior to this "release".
>
> A lot. Including systems with 2GiB RAM.
Testing on different memory sizes isn't the issue. It's testing on as
many different flavors of hardware out there that claim to be PCs.
This G41 was made by IBM, not Lenovo, so came from where it all
started. Linux, FreeBSD, OpenBSD, and NetBSD (and maybe Dragonfly --
don't recall) have all been run on this machine without
memory-size-detection issues. Just to be sure that some memory hadn't
fallen out of the machine (which I don't use much anymore), I booted
a Linux Mint Live DVD and verified that the 2 GB I thought were there
were, indeed, there.
Having said that, I do understand that a small project like this,
populated by volunteers, doesn't have access to a lot of different
examples of hardware, and apparently the G41 was among the missing. My
comment was out of frustration. This project has gone on forever, it
has tremendous potential (I'm an old OS guy -- I ran the Tenex project
at BBN and also worked on the Mach kernel at BBN; we ran it on the
Butterfly multiprocessor -- and I believe in the micro-kernel
architechture), now we finally have something resembling a release and
it can't figure out how much memory a G41 has. So I apologize a bit --
I realize what you are up against, but please understand my
frustration.
If there's anything else I can do to help debug this, let me know.
/Don
>
> Samuel
Reply to: