out_of_memory, possibly triggered by low(ish) memory system

[James Ferguson]

Binary package hint: ureadahead

Working with Lucid daily build from Apr 23rd, attempting resume from S4, as commented by cking in bug #563472

----

Looking at the dmesg recorded in comment #8 (http://launchpadlibrarian.net/45019037/dmesg-oom-boot.log):

[ 3.896224] ureadahead cpuset=/ mems_allowed=0
[ 3.896297] Pid: 413, comm: ureadahead Not tainted 2.6.32-21-generic #32-Ubuntu
[ 3.896418] Call Trace:
[ 3.896478] [<c01cc814>] oom_kill_process+0xa4/0x2b0
[ 3.896536] [<c01cce89>] ? select_bad_process+0xa9/0xe0
[ 3.896579] [<c01ccf11>] __out_of_memory+0x51/0xa0
[ 3.896620] [<c01ccfb8>] out_of_memory+0x58/0xb0
[ 3.896661] [<c01cf7c7>] __alloc_pages_slowpath+0x407/0x4a0
[ 3.896705] [<c01cf99a>] __alloc_pages_nodemask+0x13a/0x170
[ 3.900907] [<c01cf9ec>] __get_free_pages+0x1c/0x30
[ 3.904998] [<c01ad57a>] ring_buffer_resize+0x18a/0x2a0
[ 3.909094] [<c01aeb55>] tracing_resize_ring_buffer+0x25/0x80
[ 3.913273] [<c01b28b2>] tracing_entries_write+0xe2/0x160
[ 3.917561] [<c01b20c7>] ? trace_nowake_buffer_unlock_commit+0x37/0x50
[ 3.921848] [<c02f4414>] ? security_file_permission+0x14/0x20
[ 3.926112] [<c0207ad4>] ? rw_verify_area+0x64/0xe0
[ 3.930370] [<c0207bf2>] vfs_write+0xa2/0x1a0
[ 3.934573] [<c01b27d0>] ? tracing_entries_write+0x0/0x160
[ 3.938805] [<c0208512>] sys_write+0x42/0x70
[ 3.943043] [<c01033ec>] syscall_call+0x7/0xb

This is concerning - ureadahead is being killed by the out of memory killer because it's slurping up too many pages into memory for the fast boot.

This machine is reporting only 768MB of free memory, however it has 1GB memory installed. Memtest also records 768MB of memory. So, something is a bit strange. Is 256MB of memory being used for on-board integrated graphics memory or is the BIOS e820 call reporting the wrong memory size?

--- end cking quote

Note Colin was right - 256MB memory was reserved for the onboard graphics.