Panic during -CURRENT buildworld
This is on a system (my laptop): FreeBSD localhost 5.0-CURRENT FreeBSD 5.0-CURRENT #13: Sun May 27 23:44:24 PDT 2001 [EMAIL PROTECTED]:/common/C/obj/usr/src/sys/LAPTOP_30W i386 Mon May 28 07:27:59 PDT 2001 Recent CVSup activity: CVSup begin from cvsup14.freebsd.org at Sat May 26 03:47:01 PDT 2001 CVSup ended from cvsup14.freebsd.org at Sat May 26 03:52:48 PDT 2001 CVSup begin from cvsup14.freebsd.org at Sun May 27 03:47:01 PDT 2001 CVSup ended from cvsup14.freebsd.org at Sun May 27 03:53:36 PDT 2001 CVSup begin from cvsup14.freebsd.org at Mon May 28 03:47:00 PDT 2001 CVSup ended from cvsup14.freebsd.org at Mon May 28 03:53:51 PDT 2001 I had tried the buildworld within X (as had been my normal practice until the recent difficulties with swap and/or VM), and the system re-booted itself. Got the well-discussed symptom of an active file system failing fsck's check of primary vs. first alternate superblock, and after fsck got finished with the file system, soft updates got turned off, so I turned soft updates back on again. Since I was in single-user mode anyhow, I mounted the necessary file systems, issued a swapon -a, and proceeded to re-try the make buildworld ( friends) -- from within script (again, as I usually do). Here's a transcript of the first last parts of the typescript file: Script started on Mon May 28 01:21:19 2001 # mount cd /usr/src uname -a date make buildworld date make kernel KERNCONF=LAPTOP_30W date make installworld date mergemaster date sync df -k /dev/ad0s3a on / (ufs, local, soft-updates) devfs on /dev (devfs, local) /dev/ad0s3e on /usr (ufs, local, soft-updates) /dev/ad0s3g on /var (ufs, local, soft-updates) procfs on /proc (procfs, local) /dev/ad0s3h on /common (ufs, local, soft-updates) FreeBSD 5.0-CURRENT FreeBSD 5.0-CURRENT #13: Sun May 27 23:44:24 PDT 2001 [EMAIL PROTECTED]:/common/C/obj/usr/src/sys/LAPTOP_30W i386 Mon May 28 01:22:06 PDT 2001 -- Rebuilding the temporary build tree -- rm -rf /usr/obj/usr/src/i386 ... [elided -- dhw] cc -pg -O -pipe -I. -I/usr/src/lib/libncurses -I/usr/src/lib/libncurses/../../c ontrib/ncurses/ncurses -I/usr/src/lib/libncurses/../../contrib/ncurses/include - Wall -DFREEBSD_NATIVE -DNDEBUG -DHAVE_CONFIG_H -DTERMIOS -I/usr/obj/usr/src/i386 /usr/include -c /usr/src/lib/libncurses/../../contrib/ncurses/ncurses/tinfo/get env_num^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@ And a hand-transcription of the panic (starting with the last command shown on the console from the make buildworld): cc -fpic -DPIC -O -pipe -I. -I/usr/src/lib/libncurses -I/usr/src/lib/libncurses/../../contrib/ncurses/ncurses -I/usr/src/lib/libncurses/../../contrib/ncurses/include -Wall -DFREEBSD_NATIVE -DNDEBIG -DHAVE_CONFIG_H -DTERMIOS -i/usr/obj/usr/src/i386/usr/include -c lib_gen.c -o libgen.So freeing uidinfo: uid=0, proccnt=33 kernel trap 12 with interrupts disabled panic: blockable sleep lock (sleep mutex) Giant @/usr/src/sys/vm/vm_fault.c:213 Debugger(panic) Stopped atDebugger+0x44: pushl %ebx db trace Debugger(c03a499b) at Debugger+0x44 panic(c03a75e0,c03a3820,c03cc4b4,c03c1d9b,d5) at panic+0x70 witness_lock(c047dda0,8,c03c1d9b,d5) at witness_lock+0x1b2 vm_fault(c04692ac,deadc000,1,0,0) at vm_fault+0xb2 trap_pfault(ce7f4e34,0,deadc2af,ce7ffa60,c0e4259c) at trap_pfault+0x5d0 trap(ce7f0018,c01f0010,c01f0010,4,c0e4259c) at trap+0x5d0 calltrap() at calltrap+0x5 --- trap 0xc, eip=0xc01d8cb6, esp=0xce7f4e74, ebp=0xce7f4e80 --- uihold(c0e42580,c1ca3a68,c03a4235,0,98) at uihold+0x5f crdup(c0e3d600, ce7ffb7c,ce7ffa60,2,c0445a00) at crdup+0x4c access(ce7ffa60,ce7f4f80,806b240,806f080,805d1ce) at access+0x18 syscall(2f,2f,2f,805d1ce,806f080) at syscall+0x71d syscall_with_err_pushed() at syscall_with_err_pushed+0x1b In addition to the kernel.old (dated a few hours earlier the same day), I have a saved kernel from 16 May (which is the one from which I had booted in order to build the one from 23:50 (PDT) on 27 May, so I could try that. Or I could try some other things, if that might help identify the problem. I have not (yet) tried any of the posted experimental patches against anything involving file systems, soft updates, swap, or VM. (I do have a small patch for keyboard control of the sound mute function, as well as some bits pieces of some of Doug Ambrisko's sys/dev/an patches. However, the Cisco/Aironet card wasn't inserted at any point during the boot that did the panic.) Help? Thanks, david -- David H. Wolfskill [EMAIL PROTECTED] As a computing professional, I believe it would be unethical for me to advise, recommend, or support the use (save possibly for personal amusement) of any product that is or depends on any Microsoft product. To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-current in the body of the message
Re: Panic during -CURRENT buildworld
Date: Mon, 28 May 2001 10:46:53 -0700 (PDT) From: David Wolfskill [EMAIL PROTECTED] This is on a system (my laptop): FreeBSD localhost 5.0-CURRENT FreeBSD 5.0-CURRENT #13: Sun May 27 23:44:24 PDT 2001 [EMAIL PROTECTED]:/common/C/obj/usr/src/sys/LAPTOP_30W i386 Mon May 28 07:27:59 PDT 2001 Recent CVSup activity: CVSup begin from cvsup14.freebsd.org at Sat May 26 03:47:01 PDT 2001 CVSup ended from cvsup14.freebsd.org at Sat May 26 03:52:48 PDT 2001 CVSup begin from cvsup14.freebsd.org at Sun May 27 03:47:01 PDT 2001 CVSup ended from cvsup14.freebsd.org at Sun May 27 03:53:36 PDT 2001 CVSup begin from cvsup14.freebsd.org at Mon May 28 03:47:00 PDT 2001 CVSup ended from cvsup14.freebsd.org at Mon May 28 03:53:51 PDT 2001 I had tried the buildworld within X (as had been my normal practice until the recent difficulties with swap and/or VM), and the system re-booted itself. Got the well-discussed symptom of an active file system failing fsck's check of primary vs. first alternate superblock, and after fsck got finished with the file system, soft updates got turned off, so I turned soft updates back on again. I was able to do the buildworld ( friends) by booting a saved kernel from 16 May into single-user mode, so I'm now running: FreeBSD dhcp-133.catwhisker.org 5.0-CURRENT FreeBSD 5.0-CURRENT #14: Mon May 28 09:56:14 PDT 2001 root@:/common/C/obj/usr/src/sys/LAPTOP_30W i386 And while running that kernel (no further CVSups; no further source tree mods), I was able to do a make buildworld while running X. Seems like an improvement to me, david -- David H. Wolfskill [EMAIL PROTECTED] As a computing professional, I believe it would be unethical for me to advise, recommend, or support the use (save possibly for personal amusement) of any product that is or depends on any Microsoft product. To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-current in the body of the message
Re: Panic during -CURRENT buildworld
On Fri, May 18, 2001 at 09:25:59PM -0700, David Wolfskill wrote: I haven't seen anyone else reporting any problems similar to what I experienced, so I'm not about to claim there's something that's definitely broken I have seen exactly the same - the machine (IBM thinkpad T21) freezes during buildworld (or it appears to, but as you said, it's hard to say if it panic'ed when you run X). This problem appear to have been introduced sometime within the last 3-4 days. Only reason I've been silent is that I still haven't had the time to get a trace. david Cheers, /Niels Chr. -- Niels Christian Bank-Pedersen, NCB1-RIPE. Network Manager, TDC, IP-section. Hey, are any of you guys out there actually *using* RFC 2549? To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-current in the body of the message
Re: Panic during -CURRENT buildworld
Panic w/ softupdate disappears after I grab this revision of ffs_softdep.c: ident /usr/src/sys/ufs/ffs/ffs_softdep.c /usr/src/sys/ufs/ffs/ffs_softdep.c: $FreeBSD: src/sys/ufs/ffs/ffs_softdep.c,v 1.97 2001/05/19 19:24:26 mckusick Exp $ Now it's fairly smooth to buildworld, installworld, copy many small files bewteen different slice/media/network (Okay, samba :D) for me. -- Clive Lin (Tong-I Lin)\n =P [EMAIL PROTECTED] # Family, friends, private affairs\n =F [EMAIL PROTECTED] # Chinese ports, documentation\n =O [EMAIL PROTECTED] # Others\n =J.* # What do you think about the 'J' ?\n To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-current in the body of the message
Panic during -CURRENT buildworld
[This is *really* long. Sorry. dhw] Running: FreeBSD dhcp-140.catwhisker.org 5.0-CURRENT FreeBSD 5.0-CURRENT #1: Thu May 17 09:13:03 PDT 2001 [EMAIL PROTECTED]:/common/C/obj/usr/src/sys/LAPTOP_30W i386 (The #1 sequence number is probably misleading; I'll go into that below. It was around #60 or so before the events discussed there.) Last few CVSups: CVSup begin from cvsup14.freebsd.org at Tue May 15 03:47:00 PDT 2001 CVSup ended from cvsup14.freebsd.org at Tue May 15 03:52:15 PDT 2001 CVSup begin from cvsup14.freebsd.org at Wed May 16 03:47:01 PDT 2001 CVSup ended from cvsup14.freebsd.org at Wed May 16 03:52:26 PDT 2001 CVSup begin from cvsup14.freebsd.org at Wed May 16 07:57:56 PDT 2001 CVSup ended from cvsup14.freebsd.org at Wed May 16 08:04:15 PDT 2001 CVSup begin from cvsup14.freebsd.org at Thu May 17 03:47:00 PDT 2001 CVSup ended from cvsup14.freebsd.org at Thu May 17 03:52:58 PDT 2001 CVSup begin from cvsup14.freebsd.org at Fri May 18 03:47:01 PDT 2001 CVSup ended from cvsup14.freebsd.org at Fri May 18 03:52:34 PDT 2001 (Wednesday wasn't the best of days for me) As some folks may recall, I've been tracking -STABLE -CURRENT on this machine (my laptop) since early March. I've encountered various forms of challenges in that, but today's the first time I got a panic during the make buildworld. And this was on the 3rd attempt today. :-( (The other 2 were done -- as usual for me -- within an X environment, so I could more easily monitor the progress... or so I had imagined. Each of those locked up; I suspect, in hindsight, that they may also have been panics.) This last was one I did in single-user mode. (I had already built booted today's -STABLE.) FWIW, the system appears to have gotten much further while in single-user mode -- it was in stage 4: building libraries; here are my (hand-transcribed) notes: cc -fpic -DPIC ... lib_addch.So cc -fpic -DPIC ... lib_addstr.So Fatal trap 12: page fault while in kernel mode fault virtual address = 0xdeadc0e8 fault code = supervidor read, page not present instruction pointer = 0x880 0xc0314938 [This character may be wrong--^ sorry. :-( Could be 6, b, or 8.] stack pointer = 0x10: 0xccd8be7c frame pointer = 0x10: 0xccd8be7c code segment= base 0x0, limit 0xf, type 0x1b = DPL 0, pres 1, def32 1, gran 1 processor eflags= interrupt enabled, resume, IOPL=0 current process = 16 (irq14: ata0) kernel: type 12 trap, code =0 stopped atworklist_remove+0x1c: cmpw $0, 0xa(%ecx) db trace worklist_remove(deadc0de) at ++0x1c free_diradd(deadc0de) at free_diradd+0x26 free_newdirblk(c16a0980) at free_newdirblk+0x2e handle_written_inodeblock(c18a0280, c688ff6c) at handle_written_inodeblock+0x2b2 softdep_disk_write_complete(c688ff6c) at softdep_disk_write_complete+0x6a bufdone(c688ff6c, ccd8bf50, c0134df6, c688ff6c, c167c200) at bufdone+0x109 bufdonebio(c688ff6c) at bufdonebio+0xe ad_interrupt(c1be4700, c016131c0, ccd8bf7c, c01c1db7, c167c200) at ad_interrupt+0x3ce ata_intr(c167c200) at ata_intr+0xae ithread_loop(c167c180, ccd8bfa8) at ithread_loop+0x413 fork_exit(c01c19a4, c167c180, ccd8bfa8) at fork_exit+0xb4 fork_trampoline() at fork_trampoline+0x8 db Each of the first 2 times the make buildworld didn't complete, I ended up power-cycling the machine. fsck wasn't especially happy about this, and I did take the opportunity to note that each time the -CURRENT fsck wanted a hand-invoked fsck, the following effects were noted: * It complained about the primary superblock had some (presumably unexpected) discrepancy with the first alternate (at block 32). Each time, I told it to use the alternate, and to update the primary superblock. * After finishing up, the soft updates flag was no longer turned on, so I ran tunefs -n enable while I was (still) in single-user mode. As to the events alluded to above: Wednesday, while trying to figure out some breakage, I quoted the $FreeBSD$ line of a Makefile to a correspondent (Ruslan Ermilov) who was kind enough to try to help me figure things out. He noticed that the revision level on the Makefile seemed too low; off by one, actually. But cvs log Makefile in the directory in question showed that the Makefile itself was current; it was the $FreeBSD$ tag that was incorrect. It turns out that when I had set up my CVS repository, I placed it in /cvs, but put the FreeBSD part of it in /cvs/freebsd (assuming that I might want to place some non-FreeBSD-specific things in there at some point). And I set up my (default) $CVSROOT environment variable to /cvs. So far, so good. This was (as mentioned) back in early March, so there are some aspects of this that I don't quite recall. But when I created my -CURRENT working directory for /usr/src, I must have done something that did
Re: Panic during -CURRENT buildworld
Hello everybody, attention! pure speculation and unprofessional comments follow! These problems with the alternate superblock remind me... there were reports about the same when fsck had problems some time ago. But there was a common theme to all of them: The fsck raves were a whole lot more severe if there were softupdates enabled. I for example have been running -CURRENT for quite a while here on a daily basis (although I have never tried to use the same file systems concurrently with -STABLE) doing buildowrlds, Mozilla bi-daily builds and other fun stuff, yet have not seen any problems of this kind. Even when I had a crash, a manual fsck (for safety's sake) always fixed things as it should, and never complained. And this, although I had some very unfortunate crashes, when eg I crashed from X in the middle of a Mozilla build, but even so there were almost no file structures damaged or at least not noted by fsck. But I have never enabled soft updates on any fs of mine, which of course means that some operations require all the time in the world to complete, but seems it is safer somehow. I am probably just lucky and totally wrong, but just speculating... -- Regards: Szilveszter ADAM Szeged University Szeged Hungary To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-current in the body of the message
Re: Panic during -CURRENT buildworld
Date: Fri, 18 May 2001 10:48:19 -0700 (PDT) From: David Wolfskill [EMAIL PROTECTED] [Excruciatingly long narrative of panic during buildworld for today's -CURRENT elided; it's in the archives. dhw] Reporting back after getting today's -CURRENT built: FreeBSD dhcp-140.catwhisker.org 5.0-CURRENT FreeBSD 5.0-CURRENT #2: Fri May 18 11:31:46 PDT 2001 root@:/common/C/obj/usr/src/sys/LAPTOP_30W i386 Taking a cue from Szilveszter Adam's response to my note, I booted into single-user mode, turned off soft updates for the file systems on ad0s3 (all of which get used during the biuldworld/installworld process, since /usr/obj is a symlink to somwhere in /common -- df listing below for reference), and I unmounted the file systems on ad0s1 ad0s2. I then did the make buildworld/kernel/installworld mergemaster while remaining in single-user mode, running yesterday's -CURRENT (but, as noted, with soft updates turned off for all mounted file systems). As noted, it appears to have completed successfully. I have turned soft updates back on for everything. Tomorrow's build may prove interesting. Here's what df -k looks like (while I'm running -CURRENT): Filesystem 1K-blocks UsedAvail Capacity Mounted on /dev/ad0s3a 95263737821386084%/ devfs 110 100%/dev /dev/ad0s1a 95263408124683047%/S1 /dev/ad0s1e915695 7774726496892%/S1/usr /dev/ad0s2a 95263410304661247%/S2 /dev/ad0s2e915727 7761086636192%/S2/usr /dev/ad0s3e915727 733431 10903887%/usr /dev/ad0s3g254063 107557 12618146%/var /dev/ad0s3h 14116697 4180188 880717432%/common procfs 440 100%/proc /dev/md10c 520140 16 478516 0%/tmp I haven't seen anyone else reporting any problems similar to what I experienced, so I'm not about to claim there's something that's definitely broken Cheers, david -- David H. Wolfskill [EMAIL PROTECTED] As a computing professional, I believe it would be unethical for me to advise, recommend, or support the use (save possibly for personal amusement) of any product that is or depends on any Microsoft product. To Unsubscribe: send mail to [EMAIL PROTECTED] with unsubscribe freebsd-current in the body of the message