NetBSD Problem Report #24743

Received: (qmail 19911 invoked by uid 605); 11 Mar 2004 12:55:08 -0000
Message-Id: <20040311125505.9249170C21@mkd.ath.cx>
Date: Thu, 11 Mar 2004 13:55:05 +0100 (CET)
From: Claudio M.Camacho <claudiocamacho@vodafone.es>
Sender: gnats-bugs-owner@NetBSD.org
Reply-To: Claudio M.Camacho <claudiocamacho@vodafone.es>
To: gnats-bugs@gnats.NetBSD.org
Subject: LFS doesnt work with large copies
X-Send-Pr-Version: 3.95

>Number:         24743
>Category:       kern
>Synopsis:       LFS doesnt work with large copies
>Confidential:   no
>Severity:       serious
>Priority:       high
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Thu Mar 11 12:56:00 +0000 2004
>Closed-Date:    
>Last-Modified:  Sun Mar 27 02:53:00 +0000 2005
>Originator:     mkd
>Release:        NetBSD 1.6ZK
>Organization:
>Environment:
System: NetBSD Insomnia 1.6ZK NetBSD 1.6ZK (MYKERNEL) #5: Wed Mar 10 23:55:29 CET 2004 root@Insomnia:/usr/src/sys/arch/i386/compile/MYKERNEL i386
Architecture: i386
Machine: i386 (MULTIPROCESSOR, 2 Logical CPUs HTT)
>Description:
	Doing tests again using LFS filesystem, I observed that sometimes,
	LFS doesnt work, most of all when I'm trying to copy enough MBs/GBs.
	I was copying a 1GB directory into an LFS partition (in /home) and
	the copy was stopped with the next message from the LFS:

	# cp -fRP /foo /home/foo
	ino 909 bytes == 0 offset 5373952

	The syste was not stopped, not halted, just that console was
	died. I broke into the kernel DDB to see what's happening, I
	did Ctrl+Alt+Esc (i386) to see any crash but DDB didn't tell
	me about a crash, it showed:

	Stopped at netbsd:cpu_Debugger+0x4: pop %ebp
	internal_command(...) at netbsd:internal_command+0x15c
	wskbd_translate(...) at netbsd:....
	wskbd_input(...) at netbsd:...
	pckbd_input(...) at ....
	pckbcintr(...) at...
	Xintr_ioapic_edge1() at ...
	--- interrupt ---
	netbsd:cpu_switch+0x02:
	db{0}>

	I gess that above means that nothing wrong is happening, so
	I typed *continue* in the DDB and I could continue using
	my system but the console where 'cp' command was; I pressed
	again Ctrl+Alt+Esc and tried with *sync* and I got this:

	db{0}> sync
	syncing disks... kernel: page fault trap, code=0
	stopped at netbsd:lfs_seglock+0x46: movl 0x34(%ecx),%eax

>How-To-Repeat:
	Make a LFS partition, mount it, try to copy a directory
	with near 1GB or so, and you'll get the 'ino ...' message.

>Fix:
	Dont know.
>Release-Note:
>Audit-Trail:

From: Bang Jun-Young <junyoung@NetBSD.org>
To: gnats-bugs@NetBSD.org
Cc:  
Subject: Re: kern/24743
Date: Fri, 9 Jul 2004 17:07:35 +0900

 I'm suffering from the same symptom. :-(

 Jun-Young


From: Miles Nordin <carton@Ivy.NET>
To: gnats-bugs@netbsd.org
Cc: 
Subject: Re: kern/24743
Date: Sat, 26 Mar 2005 21:51:55 -0500

 I'm getting this also, with NetBSD 2.0_RC5 2005-11-14:

 ino 1280172 bytes == 0 offset 1966080

 then, processes accessing the disk are hung:

   0 16735  5979   2 -18  0    280   388 lfsput   DL   ??   8:02.63 pax -rpe 
   0  4556  6127  30 -18  0    108   428 uvn_fp2  DW   p0   0:00.24 /bin/cp /usr/pkgsrc/distfiles/postfix/postfix-2.1.5.tar.gz /usr/pkgsrc/distfiles/postfix/postfix-2.1.5.tar.gz.temp
   0 17057 22965   4  -2  0    124   668 vnlock   DW   p1   0:01.66 du -sk /mnt/proserpima-backup/ 
   0  2323 22965   0  -5  0     20   324 lfs segl D    p1   0:00.02 umount -f /mnt 
   0 28140 22965   0  -5  0     12   344 lfs segl D    p1   0:00.03 sync

 If I need LFS, should I be using -current instead of netbsd-2-0?

>Unformatted:

Home	PR Database Search