Most recent kernel where this bug did not occur: 2.6.15 Distribution: Gentoo Base System version 1.6.13 Hardware Environment: 4-way Opteron, 8G RAM, LSI Megaraid 320-2X Software Environment: Thread model: posix gcc version 3.4.4 (Gentoo 3.4.4-r1, ssp-3.4.4-1.0, pie-8.7.8) Compilation info: CC='gcc' CFLAGS='-march=opteron -mtune=opteron -O4 -pipe -D_GNU_SOURCE -DHAVE_ERRNO_AS_DEFINE=1' CXX='gcc' CXXFLAGS='-march=opteron -mtune=opteron -O4 -pipe -D_GNU_SOURCE -felide-constructors -fno-exceptions -fno-rtti -DUSE_MYSYS_NEW' LDFLAGS='' ASFLAGS='' LIBC: lrwxrwxrwx 1 root root 13 Dec 28 21:18 /lib/libc.so.6 -> libc-2.3.5.so -rwxr-xr-x 1 root root 1255872 Dec 28 21:18 /lib/libc-2.3.5.so -rw-r--r-- 1 root root 3890174 Dec 28 21:18 /usr/lib/libc.a -rwxr-xr-x 1 root root 210 Dec 28 21:18 /usr/lib/libc.so Configure command: ./configure '--prefix=/usr' '--libdir=/usr/lib' '--exec-prefix=/usr' '--datadir=/usr/share' '--libexecdir=/usr/sbin' '--sysconfdir=/etc' '--mandir=/usr/share/man' '--infodir=/usr/share/info' '--includedir=/usr/include' '--localstatedir=/var/lib/mysql' '--enable-assembler' '--with-mysql-user=mysql' '--with-innodb' '--with-ndbcluster' '--with-unix-socket-path=/var/run/mysqld/mysqld.sock' '--with-charset=cp1251' '--with-extra-charsets=latin1,koi8r,cp1251,utf8' '--with-openssl' '--without-readline' '--without-libedit' '--enable-thread-safe-client' '--with-named-thread-libs=-lpthread' '--build=x86_64-pc-linux-gnu' 'CFLAGS=-march=opteron -mtune=opteron -O4 -pipe -D_GNU_SOURCE -DHAVE_ERRNO_AS_DEFINE=1' 'CXXFLAGS=-march=opteron -mtune=opteron -O4 -pipe -D_GNU_SOURCE -felide-constructors -fno-exceptions -fno-rtti -DUSE_MYSYS_NEW' 'CXX=gcc' 'build_alias=x86_64-pc-linux-gnu' Problem Description: Stress testing of new quad-Opteron host triggers BUG at fs/inode.c:1057. It more or less stable ( during a fiew hours ) occur under heavy load caused by dbench with concurrency 500 ( dbench -c /usr/share/dbench/client_plain.txt 500 ) when dbench reads/writes on XFS partition. I've tried to reproduce it on reiserfs over the ~24 hours but with no success. [68621.791989] ----------- [cut here ] --------- [please bite here ] --------- [68621.800197] Kernel BUG at fs/inode.c:1057 [68621.804892] invalid operand: 0000 [1] SMP [68621.809729] CPU 1 [68621.812125] Modules linked in: ipt_REJECT ipt_state iptable_filter [68621.819332] Pid: 25756, comm: dbench Not tainted 2.6.15 #2 [68621.825760] RIP: 0010:[<ffffffff801844a0>] <ffffffff801844a0>{generic_delete_inode+217} [68621.834921] RSP: 0018:ffff81016af5fec8 EFLAGS: 00010202 [68621.841313] RAX: ffff81000801bfc0 RBX: ffff81012f185078 RCX: 0000000000000034 [68621.849566] RDX: 0000000000000003 RSI: ffff81012f185298 RDI: ffff81000801bfb8 [68621.857881] RBP: 0000000000000000 R08: 0000000000000003 R09: 0000000000000000 [68621.866222] R10: ffff8100dc0d1b48 R11: ffff810112178a80 R12: ffff81013324be78 [68621.874533] R13: 00007fffff8b9d80 R14: ffff81012f185078 R15: 0000000000516010 [68621.882855] FS: 00002aaaaade6ae0(0000) GS:ffffffff80448880(0000) knlGS:0000000000000000 [68621.892329] CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b [68621.898980] CR2: 00002aaaaaac1008 CR3: 000000015dd74000 CR4: 00000000000006e0 [68621.907288] Process dbench (pid: 25756, threadinfo ffff81016af5e000, task ffff8101623c5900) [68621.923377] Stack: ffff81012b857000 0000000000000000 ffff81012b857000 ffffffff8017b933 [68621.932502] ffff8100eb50ceb8 ffff8101feeb3bc0 0000000a585b4dce ffff81012b85701c [68621.941871] 0000000000000010 ffff810100000000 [68621.947755] Call Trace:<ffffffff8017b933>{sys_unlink+237} <ffffffff80185b21>{mntput_no_expire+23} [68621.958197] <ffffffff8016c163>{filp_close+89} <ffffffff8010d58e>{system_call+126} [68621.967720] [68621.970445] [68621.970446] Code: 0f 0b 68 73 61 31 80 c2 21 04 48 89 df 5b 5b 5d e9 ec eb ff [68621.980818] RIP <ffffffff801844a0>{generic_delete_inode+217} RSP <ffff81016af5fec8> Steps to reproduce: run dbench -c /usr/share/dbench/client_plain.txt 500 in circle on XFS partition over a few hours. http://sysadminday.org.ru/config-2.6.15
Is this issue still present in kernel 2.6.19?
sorry for a long delay. Original server now is unavailable for me and i couldn't reproduce it on 2x 2Core Opteron at (pretty) the same conditions with the 2.6.18-19. So i think it should be closed now.
Thanks for this information.