All of lore.kernel.org
 help / color / mirror / Atom feed
From: Mark Cave-Ayland <mark.cave-ayland@siriusit.co.uk>
To: bearcat.pilot@gmail.com, qemu-devel <qemu-devel@nongnu.org>
Subject: Re: [Qemu-devel] Sparc Softmmu
Date: Sun, 19 Feb 2012 21:31:55 +0000	[thread overview]
Message-ID: <4F416A4B.1070200@siriusit.co.uk> (raw)
In-Reply-To: <4F411906.3020902@gmail.com>

On 19/02/12 15:45, P. Wilhelm wrote:

Hi Paul,

> I've been able to install Solaris 8 using CDs on the Sparc Softmmu
> client system. Kudos to those responsible for Sparc development!
>
> I've been able to run a number of applications without problems on the
> client machine. I noticed something odd, however, and have been trying
> to isolate the cause. Hopefully, someone here will have an idea or two
> for me to try.
>
> The issue:
> The syslogd seems to accept and post to the appropriate log file only a
> small number of messages before no longer updating the log file when
> further messages are posted, the syslogd seems to hang. The symptom does
> not appear to be different when rebooting or restarting the syslog
> daemon. The daemon will post a couple of message to the log file and
> then stop accepting any more.
>
> Why ask here?
> I've done a couple of things to see if I can isolate the source of the
> oddity and they seem to point to qemu.

It does seem that you've found some kind of emulation bug here...

> What I've done so far:
> 1) I've tried using "logger" and a C program I wrote to use the syslog()
> function. - Both have the same issue noted above.

Okay - if you have a reproducible test case then this is a very good 
starting point :)

> 2) I've used both the OpenBios and SS5.bin bios. - Symptom does not
> change between the two.

Right. I don't think this would make much difference, since 
OpenBIOS/SS5.bin are mainly used to provide enough support for the 
Solaris kernel to enumerate and interface with the virtual hardware. 
Once the main Solaris kernel is running, the native kernel drivers take 
over and so OpenBIOS/SS5.bin are no longer really in the picture from 
this point onwards.

> 3) I checked my /etc/syslog.conf on real hardware running the same
> version of Solaris 8. Syslogging works as you'd expect there. (Note - I
> don't have real SparcStation 5 hardware. I've been using an old Sun4u
> machine, Ultra-1 -- hopefully, that does not invalidate my "real
> hardware" checks.).

Okay - I'm not sure about this, but Blue/Artyom would probably be better 
to comment here.

> 4) I ran syslogd in debug mode on both the client and the real hardware,
> but did not see anything in the output from each that gave a clue as to
> the issue. Generally, the output confirmed that I had syslogd configured
> the same way on both.
>
> How to proceed?
> I am a reasonably adept software developer, however, I do not have
> experience at the guts-level of Solaris OS or Sparc hardware. My work on
> Solaris/Sparc has been at the application level, but I have worked at
> the hardware level on other (proprietary) systems. If I had access to
> syslogd source code, I'd be comfortable working from there, but I am
> fairly certain that is not available - let me know if I am wrong. I've
> thought about looking for an open source syslog daemon and trying to use
> it instead of the Solaris version.
>
> Any thoughts about next steps are appreciated.

If you head on over to the OpenSolaris source repository then you can 
grab a copy of Oracle's syslogd code from 
http://src.opensolaris.org/source/xref/onnv/onnv-gate/usr/src/cmd/syslogd/. 
It's probably a much later version than the one supplied with Solaris 8, 
however if you can reproduce the problem using source code from there 
then we can definitely point you in the right direction to help find out 
what is wrong.

Also Blue has commented in the past that because it's only recently been 
possible to run SPARC Solaris under QEMU, it could be that the Solaris C 
compiler emits different instruction sequences to, say, GCC, and so it 
may be testing instruction paths in the emulator that aren't typically 
used by GCC/Linux which I believe was one of the primary development 
platforms. So you may like to experiment building your syslogd with both 
GCC and the native Solaris CC and then comparing the results to see if 
this makes a difference too.


HTH,

Mark.

-- 
Mark Cave-Ayland - Senior Technical Architect
PostgreSQL - PostGIS
Sirius Corporation plc - control through freedom
http://www.siriusit.co.uk
t: +44 870 608 0063

Sirius Labs: http://www.siriusit.co.uk/labs

  reply	other threads:[~2012-02-19 21:32 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2012-02-19 15:45 [Qemu-devel] Sparc Softmmu P. Wilhelm
2012-02-19 21:31 ` Mark Cave-Ayland [this message]
     [not found] ` <CACXAS8AT_WeKBNxyj+=whhXoUNe7QpCPWDTRDGDnjrqu+SbO2Q@mail.gmail.com>
2012-02-23 23:47   ` [Qemu-devel] [offtopic] " P. Wilhelm
2012-02-26 21:34     ` Blue Swirl

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=4F416A4B.1070200@siriusit.co.uk \
    --to=mark.cave-ayland@siriusit.co.uk \
    --cc=bearcat.pilot@gmail.com \
    --cc=qemu-devel@nongnu.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.