I was running "top" and it stopped updating. At the end of this message is top's trace out of a sysrq-t. I don't understand why a file *read* resulted in waiting for a write to complete. At a guess, the read attached itself to a transaction that was in the process of write-out? Would it be possible for a read to return the data without waiting for the writes to actually reach disk? top D ffff810003dfd580 0 12032 12013 ffff810013c2d988 0000000000000092 0000000000000292 ffffffff803c2b09 ffff81001044b7c8 ffffffff80918160 0000000000000002 ffffffff807ff700 ffffffff807ff700 ffffffff807ff700 ffffffff807ff700 ffffffff807ff700 Call Trace: [<ffffffff803c2b09>] cfq_set_request+0xe9/0x3d0 [<ffffffff80258f36>] getnstimeofday+0x36/0xb0 [<ffffffff8053433d>] io_schedule+0x5d/0xa0 [<ffffffff803b84b9>] get_request_wait+0xe9/0x190 [<ffffffff80254060>] autoremove_wake_function+0x0/0x30 [<ffffffff803b9a90>] __make_request+0x90/0x660 [<ffffffff802a0000>] add_to_swap+0x70/0x90 [<ffffffff803b68f3>] generic_make_request+0x163/0x320 [<ffffffff8025d847>] mark_held_locks+0x57/0xa0 [<ffffffff80291cc5>] inc_zone_page_state+0x65/0x80 [<ffffffff803b94cc>] submit_bio+0x8c/0x130 [<ffffffff8033d08d>] write_jnodes_to_disk_extent+0x40d/0x440 [<ffffffff8033d175>] write_jnode_list+0xb5/0xf0 [<ffffffff8033e3d3>] reiser4_write_logs+0x733/0xda0 [<ffffffff803378bd>] reiser4_txn_end+0x3ad/0x570 [<ffffffff8025da01>] trace_hardirqs_on+0xc1/0x160 [<ffffffff80337a89>] reiser4_txn_restart+0x9/0x20 [<ffffffff803509b5>] read_unix_file+0x345/0x4c0 [<ffffffff8053717f>] _spin_unlock_irq+0x2f/0x50 [<ffffffff802b0085>] vfs_read+0xc5/0x180 [<ffffffff802b0583>] sys_read+0x53/0x90 [<ffffffff8020c1de>] system_call+0x7e/0x83 -- Zan Lynx <zlynx@xxxxxxx>
Attachment:
signature.asc
Description: This is a digitally signed message part
