Hello,
unfortunately I got some trouble using ext3 - the system hangs when
working in a certain directory. Not completely, I can switch between
consoles and reboot with SysRQ, but cannot do anything which requires
disk IO: not start any applications, not umount, nothing. I could send
the strace output, but I don't think this would help anyone. Sorry, but
I don't have a clue where to start debugging. There are no
debugging messages on my syslog console in the time of the freeze when I
compile with JBD and enable via 1>/proc/sys/fs/jbd-debug.
The filesystem seems to be clean, e2fsck passes without errors. Please
let me know if (and how) I could help on debugging a such problem.
Filesystem 1k-blocks Used Available Use% Mounted on
/dev/hda7 5537490 4781331 412049 93% /
Filesystem Inodes IUsed IFree IUse% Mounted on
/dev/hda7 1441256 255539 1185717 18% /
[pid 2232] time(NULL) = 1001184942
[pid 2232] time([1001184942]) = 1001184942
[pid 2232] getpid() = 2232
[pid 2232] rt_sigaction(SIGPIPE, {0x401dfe08, [], 0x4000000}, {SIG_DFL}, 8)
0[pid 2232] send(3, "<60>Sep 22 20:55:42 makehistory["..., 106,
0) = 106
[pid 2232] rt_sigaction(SIGPIPE, {SIG_DFL}, NULL, 8) = 0
[pid 2232] fcntl64(4, F_SETLKW, {type=F_WRLCK, whence=SEEK_SET, start=66184,
len=64}) = 0
[pid 2232]
unlink("/var/spool/news/overview/d/c/o/u/l/m/de.comp.os.unix.linux.misc-NEW.IDX")
= 0
[pid 2232]
open("/var/spool/news/overview/d/c/o/u/l/m/de.comp.os.unix.linux.misc.DAT",
O_RDWR|O_APPEND|O_CREAT, 0660) = 14
[pid 2232]
open("/var/spool/news/overview/d/c/o/u/l/m/de.comp.os.unix.linux.misc.IDX",
O_RDWR|O_CREAT, 0660) = 15
[pid 2232] fstat64(15, {st_mode=S_IFREG|0640, st_size=21312, ...}) = 0
[pid 2232] fcntl64(14, F_GETFD) = 0
[pid 2232] fcntl64(14, F_SETFD, FD_CLOEXEC) = 0
[pid 2232] fcntl64(15, F_GETFD) = 0
[pid 2232] fcntl64(15, F_SETFD, FD_CLOEXEC) = 0
[pid 2232] fstat64(14, {st_mode=S_IFREG|0640, st_size=747, ...}) = 0
[pid 2232] fstat64(15, {st_mode=S_IFREG|0640, st_size=21312, ...}) = 0
[pid 2232] old_mmap(NULL, 747, PROT_READ, MAP_SHARED, 14, 0) = 0x40016000
[pid 2232] old_mmap(NULL, 21312, PROT_READ, MAP_SHARED, 15, 0) = 0x40017000
[pid 2232]
open("/var/spool/news/overview/d/c/o/u/l/m/de.comp.os.unix.linux.misc-NEW.IDX",
O_RDWR|O_CREAT, 0660) = 16
[pid 2232] fstat64(15, {st_mode=S_IFREG|0640, st_size=21312, ...}) = 0
[pid 2232] pwrite(16,
"\0\0\0\0\36\1\0\0\312v\247;\0\0\0\0\5\1\0\0\0\f\0\0\26"..., 20196,
32868) 16284
[pid 2232] time([1001184942]) = 1001184942
[pid 2232] getpid() = 2232
[pid 2232] rt_sigaction(SIGPIPE, {0x401dfe08, [], 0x4000000}, {SIG_DFL}, 8)
0[pid 2232] send(3, "<59>Sep 22 20:55:42 makehistory["..., 174,
0) = 174
[pid 2232] rt_sigaction(SIGPIPE, {SIG_DFL}, NULL, 8) = 0
[pid 2232] close(16) = 0
[pid 2232] close(15) = 0
[pid 2232] close(14) = 0
[pid 2232] munmap(0x40017000, 21312) = 0
[pid 2232] munmap(0x40016000, 747) = 0
[pid 2232] fcntl64(4, F_SETLKW, {type=F_UNLCK, whence=SEEK_SET, start=66184,
len=64}) = 0
[pid 2232] write(2, "makehistory: Can\'t write overvie"...,
80makehistory:
Can't write overview data "@05010000000B000059410000000000000000@"
) = 80
[pid 2232] time(NULL) = 1001184942
[pid 2232] time([1001184942]) = 1001184942
[pid 2232] getpid() = 2232
[pid 2232] rt_sigaction(SIGPIPE, {0x401dfe08, [], 0x4000000}, {SIG_DFL}, 8)
0[pid 2232] send(3, "<60>Sep 22 20:55:42 makehistory["..., 106,
0
Gruss/Regards,
Eduard.
--
Accepted Languages: [en, de, C]
What version of the kernel and ext3 are you using? - Ted
#include <hallo.h> Theodore Tso wrote on Sat Sep 22, 2001 um 05:40:10PM:> What version of the kernel and ext3 are you using?Oh sorry, forgot the most important thing ;) Kernel 2.4.9 build from Debian source (with various fixes). Ext3 is 0.9.9 from the Debian package (the 2.4.10pre4 version with the obligatory change to apply on 2.4.9 and your readdir speedup patch). It is a i386 machine, Athlon (with Athlon optimisation in kernel). I tried without the readdir patch - did not help. The same dead lock. Gruss/Regards, Eduard. -- #exclude <windows.h>
Eduard Bloch wrote:> > Hello, > > unfortunately I got some trouble using ext3 - the system hangs when > working in a certain directory.Well that's cute. Is it possible to test the same directory with an ext2 mount? Can you suggest any way in which this can be reproduced by others? Thanks.
> Filesystem 1k-blocks Used Available Use% Mounted on > /dev/hda7 5537490 4781331 412049 93% / >smells like an very old general problem with 93% ... i.t
Hi, On Sat, Sep 22, 2001 at 10:12:55PM +0200, Eduard Bloch wrote:> unfortunately I got some trouble using ext3 - the system hangs when > working in a certain directory. Not completely, I can switch between > consoles and reboot with SysRQ, but cannot do anything which requires > disk IO: not start any applications, not umount, nothing. I could send > the strace output, but I don't think this would help anyone. Sorry, but > I don't have a clue where to start debugging. There are no > debugging messages on my syslog console in the time of the freeze when I > compile with JBD and enable via 1>/proc/sys/fs/jbd-debug. > The filesystem seems to be clean, e2fsck passes without errors. Please > let me know if (and how) I could help on debugging a such problem.The real question is, where are the processes stuck? Can you set up a serial console? If so, please enable magic-sysrq and try to capture an alt-sysrq-T trace to find out where all the processes are blocked. That, with the System.map, will give us somewhere to start looking. Cheers, Stephen
#include <hallo.h> Stephen C. Tweedie wrote on Mon Sep 24, 2001 um 02:43:41PM:> > compile with JBD and enable via 1>/proc/sys/fs/jbd-debug. > > The filesystem seems to be clean, e2fsck passes without errors. Please > > let me know if (and how) I could help on debugging a such problem. > > The real question is, where are the processes stuck? Can you set up aGood news. Ext3 0.9.10 for 2.4.10 does not hang the system. Still no idea what happened there, but this seems to be fixed in 2.4.10. Gruss/Regards, Eduard. -- Die meisten Menschen pflegen im Kindesalter vom Zeigen auf Gegenstände (Mausbewegung) und "ga" sagen (Mausklick) abzukommen, zugunsten eines mächtigeren und langwierig zu erlernenden Tools (Sprache). Achim Linder in de.comp.os.linux.misc
Hi, On Tue, Sep 25, 2001 at 01:51:47AM +0200, Eduard Bloch wrote:> #include <hallo.h> > Stephen C. Tweedie wrote on Mon Sep 24, 2001 um 02:43:41PM: > > > compile with JBD and enable via 1>/proc/sys/fs/jbd-debug. > > > The filesystem seems to be clean, e2fsck passes without errors. Please > > > let me know if (and how) I could help on debugging a such problem. > > > > The real question is, where are the processes stuck? Can you set up a > > Good news. Ext3 0.9.10 for 2.4.10 does not hang the system. Still no > idea what happened there, but this seems to be fixed in 2.4.10.Thanks for the report. Cheers, Stephen