[9fans] dd-induced hangs

[9fans] dd-induced hangs

[andrey mirtchovski]

doing 'dd -if /dev/sdC0/arenas -of /dev/sdD0/arenas -bs 8192' hangs the
system with 100% reproducibility (though each occurs after a different
period of time).

the arenas partition is 60GB. running the same command on smaller
partitions (fossil, 13GB; isect, 3GB; 9fat, 10MB; nvram, 512B) completes
successfully (at 15MB/s transfer rates).

none of the processes on the machine are able to do any work, the console is
updated when I hit a key (interrupts are processed), but rc doesn't get a
chance to print a prompt.

the load on the machine before the event wasn't high, but stats showed it
peaking at 8000 just before the machine hung.

things like ^T^Tp work, and in fact I've captured ^T^Tp, ^T^Tq and ^T^Tx if
anyone is interested (^T^Ts doesn't work anymore? it's in the man page).
starting the debugger (^T^TD) caused the machine to hang completely and
irreversibly.

it smells like a race, but my nose isn't that good to be absolutely sure :)

andrey

ps: i've turned dma off now and am runing the same thing again, but dma off
means i'll be able to verify whether the command completed early tomorrow
morning.

[9fans] Re: dd-induced hangs

[andrey mirtchovski]

the load jumped to 6, not 8 thousand. the reason was that the ready queue at
this time had 6 processes waiting in it -- they piled up immediately after
'the event'...

see, i'm reading it and almost understanding it :) the biggest almost in the
world ;)

Re: [9fans] dd-induced hangs

[David Presotto]

[-- Attachment #1: Type: text/plain, Size: 62 bytes --]

smells like one to me also.  We should be able to recreate it.

[-- Attachment #2: Type: message/rfc822, Size: 3283 bytes --]

From: andrey mirtchovski <mirtchov@cpsc.ucalgary.ca>
To: 9fans@cse.psu.edu
Subject: [9fans] dd-induced hangs
Date: Mon, 7 Jul 2003 16:05:59 -0600 (MDT)
Message-ID: <Pine.LNX.4.44.0307071550240.3842-100000@fbsd.cpsc.ucalgary.ca>

doing 'dd -if /dev/sdC0/arenas -of /dev/sdD0/arenas -bs 8192' hangs the
system with 100% reproducibility (though each occurs after a different
period of time).

the arenas partition is 60GB. running the same command on smaller
partitions (fossil, 13GB; isect, 3GB; 9fat, 10MB; nvram, 512B) completes
successfully (at 15MB/s transfer rates).

none of the processes on the machine are able to do any work, the console is
updated when I hit a key (interrupts are processed), but rc doesn't get a
chance to print a prompt.

the load on the machine before the event wasn't high, but stats showed it
peaking at 8000 just before the machine hung.

things like ^T^Tp work, and in fact I've captured ^T^Tp, ^T^Tq and ^T^Tx if
anyone is interested (^T^Ts doesn't work anymore? it's in the man page).
starting the debugger (^T^TD) caused the machine to hang completely and
irreversibly.

it smells like a race, but my nose isn't that good to be absolutely sure :)

andrey

ps: i've turned dma off now and am runing the same thing again, but dma off
means i'll be able to verify whether the command completed early tomorrow
morning.