Try to ./configure php-5.2.8 on a reiserfs partition formated with 'mkreiserfs -b 1024' and mounted as 'notai,noatime'. Doing this as root will *hang* the system on this configure step: "checking whether pwrite works..." Watch the load average grow. Attached is what I was able to catch. I also had similar problems after rebooting withoud doing anything. On an ext3 partition or a reiserfs partition formated with a 4096 block size and it worked fine. Regards Reproducible: Always Steps to Reproduce: 1.'mkreiserfs -b 1024 /dev/sdX' 2.download php-5.2.8, untar, uncompress and run as non-root 3../configure --prefix=/srv ... Actual Results: if configures was executed as root, the machine goes down as a user, it's impossible to kill the process. Expected Results: weird things ... [<ffffffff80271ea6>] ? generic_file_aio_write+0x64/0xc4 [<ffffffff80299851>] ? do_sync_write+0xce/0x113 [<ffffffff80220a41>] ? do_page_fault+0x467/0x81a [<ffffffff802431f2>] ? autoremove_wake_function+0x0/0x2e [<ffffffff8029a126>] ? vfs_write+0xad/0x156 [<ffffffff8029a226>] ? sys_pwrite64+0x57/0x77 [<ffffffff8050b609>] ? error_exit+0x0/0x51 [<ffffffff8020b34b>] ? system_call_fastpath+0x16/0x1b Code: 44 29 eb 4a 8d 7c 2d 00 e8 5e b8 05 00 48 63 74 24 0c 48 8b 54 24 10 48 ff c6 48 89 ef 48 6b f6 18 4c 89 e9 48 03 72 28 45 85 e4 <f3> a4 66 89 5c 24 2e 48 8b 44 24 20 66 89 58 04 75 0b 48 8b 54 RIP [<ffffffff802f1644>] leaf_copy_items_entirely+0xaf/0x20f RSP <ffff880071e33328> CR2: ffff87f5a7ad9cd0 ---[ end trace 16236b432a58e107 ]--- ------------[ cut here ]------------ WARNING: at kernel/exit.c:1001 do_exit+0x3c/0x807() Modules linked in: snd_pcm_oss snd_mixer_oss snd_seq_dummy snd_seq_oss snd_seq_midi_event snd_seq snd_seq_device fuse ext3 jbd mbcache coretemp hwmon slhc realtek i2c_dev fglrx(P) snd_hda_intel snd_pcm snd_timer snd_page_alloc snd_hwdep snd e1000e 8139too soundcore i2c_i801 Pid: 24078, comm: conftest Tainted: P D 2.6.27-gentoo-r5 #5 Call Trace: [<ffffffff802314dd>] warn_on_slowpath+0x51/0x77 [<ffffffff80227f36>] enqueue_task_fair+0x22/0x74 [<ffffffff805092ee>] printk+0x4e/0x58 [<ffffffff805092ee>] printk+0x4e/0x58 [<ffffffff80234454>] do_exit+0x3c/0x807 [<ffffffff802275b8>] __wake_up+0x38/0x4e [<ffffffff8020c919>] oops_begin+0x0/0x86 [<ffffffff80220d28>] do_page_fault+0x74e/0x81a [<ffffffff8027bb78>] zone_statistics+0x3c/0x5f [<ffffffff80273b95>] __rmqueue+0x1d/0x1fb [<ffffffff802b9cd5>] __find_get_block+0x14c/0x15c [<ffffffff8050b609>] error_exit+0x0/0x51 [<ffffffff802f1644>] leaf_copy_items_entirely+0xaf/0x20f [<ffffffff802f1626>] leaf_copy_items_entirely+0x91/0x20f [<ffffffff802f31e4>] leaf_move_items+0x2ed/0x34e [<ffffffff802e2e94>] do_balance+0x2126/0x27d6 [<ffffffff802431e1>] wake_up_bit+0x11/0x22 [<ffffffff802f8462>] reiserfs_prepare_for_journal+0x5d/0x64 [<ffffffff802ed6b4>] fix_nodes+0xe0d/0x1071 [<ffffffff8050a2f3>] __wait_on_bit+0x62/0x70 [<ffffffff803dd7ff>] scsi_request_fn+0x0/0x395 [<ffffffff802f637c>] reiserfs_paste_into_item+0x101/0x13c [<ffffffff802e8f11>] reiserfs_get_block+0xd4f/0xf0f [<ffffffff802bb27e>] __block_prepare_write+0x1b3/0x3af [<ffffffff802e81c2>] reiserfs_get_block+0x0/0xf0f [<ffffffff802bb5cb>] block_write_begin+0x77/0xc5 [<ffffffff802e641f>] reiserfs_write_begin+0x145/0x1c7 [<ffffffff802e81c2>] reiserfs_get_block+0x0/0xf0f [<ffffffff80270d33>] generic_file_buffered_write+0x146/0x65d [<ffffffff802355fa>] current_fs_time+0x1e/0x24 [<ffffffff8026f8c9>] file_remove_suid+0x1c/0x4f [<ffffffff80271747>] __generic_file_aio_write_nolock+0x349/0x37d [<ffffffff80271ea6>] generic_file_aio_write+0x64/0xc4 [<ffffffff80299851>] do_sync_write+0xce/0x113 [<ffffffff80220a41>] do_page_fault+0x467/0x81a [<ffffffff802431f2>] autoremove_wake_function+0x0/0x2e [<ffffffff8029a126>] vfs_write+0xad/0x156 [<ffffffff8029a226>] sys_pwrite64+0x57/0x77 [<ffffffff8050b609>] error_exit+0x0/0x51 [<ffffffff8020b34b>] system_call_fastpath+0x16/0x1b ---[ end trace 16236b432a58e107 ]---
Do you want to try to reproduce this bug without fglrx loaded? Most kernel hackers don't like to try to debug stuff with tainted kernels.
(In reply to comment #1) This is what I got from kernel hackers: > Hello. > Yes, this is a known problem: > http://bugzilla.kernel.org/show_bug.cgi?id=9108 > > We'll try to fix it. > Please, use default blocksize for now.