home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!dtix!darwin.sura.net!sgiblab!sgigate!odin!maui.corp.sgi.com!scottm
- From: scottm@maui.corp.sgi.com (Scott Machtmes)
- Newsgroups: comp.sys.sgi.bugs
- Subject: Re: panic faulting when out of memory
- Message-ID: <1993Jan6.182531.11501@odin.corp.sgi.com>
- Date: 6 Jan 93 18:25:31 GMT
- References: <Jan05.220529.70886@yuma.ACNS.ColoState.EDU> <uh43lv0@zuni.esd.sgi.com>
- Sender: news@odin.corp.sgi.com (Net News)
- Distribution: usa
- Organization: Silicon Graphics, Inc.
- Lines: 52
- Nntp-Posting-Host: maui.corp.sgi.com
-
- In article <uh43lv0@zuni.esd.sgi.com>, olson@anchor.esd.sgi.com (Dave Olson) writes:
- |> In <Jan05.220529.70886@yuma.ACNS.ColoState.EDU> jamison@yuma.ACNS.ColoState.EDU (Jamie Gulden) writes:
- |>
- |>
- |> | This probably isn't a bug but...
- |> |
- |> | Does IRIX (4.0.4) normally panic fault if it runs out of swap
- |> | space? I've got a short 5-7 line program that I try to find
- |> | out how much memory I can allocate and it keeps panic faulting.
- |> |
- |> | The syslog has a bunch of these:
- |> |
- |> | Jan 5 13:52:35 nagel unix: dks0d3s1: Swap out failed blkno 0x3C514 (page still in memory)
- |> | Jan 5 13:52:35 nagel unix:
- |> | Jan 5 13:52:35 nagel unix: dks0d3s1: Swap out failed blkno 0x3C834 (page still in memory)
- |> | Jan 5 13:52:35 nagel unix:
- |> | Jan 5 13:52:35 nagel unix: dks0d3s1: Swap out failed blkno 0x3CA9C (page still in memory)
- |> | Jan 5 13:56:53 nagel syslogd: restart
- |> |
- |> |
- |> | And then the system comes up with this:
- |> |
- |> | Jan 5 13:57:04 nagel syslogd: restart
- |> | Jan 5 13:57:05 nagel savecore: reboot after panic:
- |> | Jan 5 13:57:05 nagel PANIC: swapoutpts - i/o error in swap
- |> |
- |> |
- |> | So, is there a disk I/O problem, did we tell it to use more blocks
- |> | than there actually are, or something else I'm missing.
- |>
- |> It isn't an out of swap problem at all. There is a disk error
- |> of some kind, *or* you have somehow configured the system to believe
- |> that there is more swap space than will fit on that partition.
- |>
- |> There are a couple of places in the swapping code where the system
- |> is fully committed, and can't recover on an error; in most cases it
- |> will just kill the process that it was trying to swap (or retry it
- |> later).
- |> --
-
- I also had a customer that had added more swap using the " swap -a " command and
- had set the swap boundary's slightly overlapping on another partition causing
- the very same symptoms.
-
- ----------------------------------------------------------------
- Scott Machtmes, System Support Engineer, Mt.View,CA,Silicon Graphics.
- Internet : scottm@sgi.com
- UUCP : {sun,decwrl,pyramid,ucbvax}!sgi!scottm
- Telephone : (415) 390-3913
- US Mail : 2171 Landings Drive, Mountain View, CA 94043
- FAX : (415) 960-3393
- SGI office mail : Mail Stop DWR-275
-