Symmetric Multiprocessing (SMP) development
 help / color / mirror / Atom feed
From: "Randy.Dunlap" <rddunlap@osdl.org>
To: "Earle R. Nietzel" <nietzel@rhinobox.org>
Cc: linux-smp@vger.kernel.org
Subject: Re: How does an UNEXPECTED_IO_APIC happen?
Date: Thu, 17 Apr 2003 10:16:03 -0700	[thread overview]
Message-ID: <20030417101603.060babc2.rddunlap@osdl.org> (raw)
In-Reply-To: <1050584281.3616.34.camel@home>

Hi Earle,

Good job.  This should be added to the linux-smp FAQ, but I can't
find it's URL just now.
I've asked the (latest) maintainer about its location...if any.


I have a few corrections below if you want them.


On 17 Apr 2003 14:58:01 +0200 "Earle R. Nietzel" <nietzel@rhinobox.org> wrote:

| How does an UNEXPECTED_IO_APIC happen?
| (as of 2.4.20)
| 
| Probably everyone has seen this common message among who has new
*                                                among those who have new
| hardware or hardware that is not very common.
|  
| "An unexpected IO-APIC was found."
| 
| The most common questions that are askes as a result are:
*                                    asked
| Many people will ask what does this mean? Is something broke? Is my
| system not stable because of this? Does this affect my performance? How
| can I fix this?
| 
| To answer all of those questions with just one answer:
| This message does not identify that something is broke, or that a system
*                                                  broken,
| is unstable, or that it could affect your performance what it does do is
*                                           performance. What
| allow kernel programmers (most likely Ingo Molnar) to see what types of
| hardware (motherboards) are out there and to create a known list (the
| vanilla list).
*         or white list).
| 
| How does your hardware get onto the list?
| That I do not know and the maintainer (Ingo Molnar) would be the person
| to ask.
[I have added some entries to the white list.  Probably Alan has also.
Probably MKP(?) also.  For me it's just a matter of having hard data,
preferably chipset specs, to validate/verify the devices.]

| How does an UNEXPECTED_IO_APIC happen?
| 
| The IO_APIC is read into 3 structures that are each 32 bits (u32) and
| are packed with different fields:
| 
| 
| register#00 = 0x02000000
|                 |||-----
|                 rIr
|                 eDe
|                 s s
|                 1 2
| 
| res1 is 4 bits (reserved)
| ID is 4 bits
| res2 is the remaining 24 bits (reserved)
| 
| 
| register#01 = 0x00178020
|                 |-|-|-|-
|                 r e Prv
|                 e n Ree
|                 s t Qsr
|                 1 r  2s
|                   i   i
|                   e   o
|                   s   n
|                       
|  res1 is 8 bits (reserved)
|  entries is 8 bits
| *PRQ is 1 bit (the 16th bit)
*               (bit 15)
[since the bits are numbered from the right/low-order side or end
and begin with 0]
| *res2 is 7 bits (reserved)
|  version is 8 bits
| 
| (*PRQ and reserved make up 8 bits with PRQ being the first)
| 
| 
| register#02 = 0x00000000
|                 |||-----
|                 rar
|                 ere
|                 sbs
|                 1i2
|                  t 
|                  r 
|                  a 
|                  t 
|                  i 
|                  o  
|                  n  
| 
| res1 is 4 bits (reserved)
| arbitration is 4 bits
| res2 is the remaining 24 bits (reserved)
| 
| 
| After reading each IO_APIC they are examined, for example:
| On my SuperMicro X5DAE (Dual Xeon 2.4G, Intel E7505):
| 
| IO APIC #2......
| .... register #00: 02000000
| .......    : physical APIC id: 02
| .... register #01: 00178020
| .......     : max redirection entries: 0017
| .......     : PRQ implemented: 1
| .......     : IO APIC version: 0020
| .... register #02: 00000000
| .......     : arbitration: 00
...
| 
| The rules for generating an UNEXPECTED_IO_APIC are:
| 
| 1)If any value other than 0 is found in the reserved bit areas (res1 ||
| res2) of the 3 registers it is considered an unexpected IO_APIC. 
| 
| 2)In register#01 if the entries field contains any other value except:
*                                                any value other than:
| 0x0f (older (Neptune) boards)
| 0x17 (typical ISA+PCI boards)
| 0x1b (Compaq Proliant boards)
| 0x1f (dual Xeon boards)
| 0x22 (bigger Xeon boards)
| 0x2E (?)
| 0x3F (?)
| 
| 3)In register#01 if the version field contains any other value except:
*                                                any value other than:
| 0x01 (82489DX IO-APICs)
| 0x02 (VIA)
| 0x10 (oldest IO-APICs)
| 0x11 (Pentium/Pro IO-APICs)
| 0x13 (Xeon IO-APICs)
| 0x20 (Intel P64H [82806 AA])
| 
| -- 

--
~Randy

      reply	other threads:[~2003-04-17 17:16 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2003-04-17 12:58 How does an UNEXPECTED_IO_APIC happen? Earle R. Nietzel
2003-04-17 17:16 ` Randy.Dunlap [this message]

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=20030417101603.060babc2.rddunlap@osdl.org \
    --to=rddunlap@osdl.org \
    --cc=linux-smp@vger.kernel.org \
    --cc=nietzel@rhinobox.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox