Revision fb59e9f1e9786635ea12e12bf6adbb132e10f979 authored by Hugh Dickins on 04 March 2008, 22:29:16 UTC, committed by Linus Torvalds on 05 March 2008, 00:35:15 UTC
While testing force_empty, during an exit_mmap, __mem_cgroup_remove_list called from mem_cgroup_uncharge_page oopsed on a NULL pointer in the lru list. I couldn't see what racing tasks on other cpus were doing, but surmise that another must have been in mem_cgroup_charge_common on the same page, between its unlock_page_cgroup and spin_lock_irqsave near done (thanks to that kzalloc which I'd almost changed to a kmalloc). Normally such a race cannot happen, the ref_cnt prevents it, the final uncharge cannot race with the initial charge. But force_empty buggers the ref_cnt, that's what it's all about; and thereafter forced pages are vulnerable to races such as this (just think of a shared page also mapped into an mm of another mem_cgroup than that just emptied). And remain vulnerable until they're freed indefinitely later. This patch just fixes the oops by moving the unlock_page_cgroups down below adding to and removing from the list (only possible given the previous patch); and while we're at it, we might as well make it an invariant that page->page_cgroup is always set while pc is on lru. But this behaviour of force_empty seems highly unsatisfactory to me: why have a ref_cnt if we always have to cope with it being violated (as in the earlier page migration patch). We may prefer force_empty to move pages to an orphan mem_cgroup (could be the root, but better not), from which other cgroups could recover them; we might need to reverse the locking again; but no time now for such concerns. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1 parent 9b3c0a0
Kconfig.preempt
choice
prompt "Preemption Model"
default PREEMPT_NONE
config PREEMPT_NONE
bool "No Forced Preemption (Server)"
help
This is the traditional Linux preemption model, geared towards
throughput. It will still provide good latencies most of the
time, but there are no guarantees and occasional longer delays
are possible.
Select this option if you are building a kernel for a server or
scientific/computation system, or if you want to maximize the
raw processing power of the kernel, irrespective of scheduling
latencies.
config PREEMPT_VOLUNTARY
bool "Voluntary Kernel Preemption (Desktop)"
help
This option reduces the latency of the kernel by adding more
"explicit preemption points" to the kernel code. These new
preemption points have been selected to reduce the maximum
latency of rescheduling, providing faster application reactions,
at the cost of slightly lower throughput.
This allows reaction to interactive events by allowing a
low priority process to voluntarily preempt itself even if it
is in kernel mode executing a system call. This allows
applications to run more 'smoothly' even when the system is
under load.
Select this if you are building a kernel for a desktop system.
config PREEMPT
bool "Preemptible Kernel (Low-Latency Desktop)"
help
This option reduces the latency of the kernel by making
all kernel code (that is not executing in a critical section)
preemptible. This allows reaction to interactive events by
permitting a low priority process to be preempted involuntarily
even if it is in kernel mode executing a system call and would
otherwise not be about to reach a natural preemption point.
This allows applications to run more 'smoothly' even when the
system is under load, at the cost of slightly lower throughput
and a slight runtime overhead to kernel code.
Select this if you are building a kernel for a desktop or
embedded system with latency requirements in the milliseconds
range.
endchoice
config RCU_TRACE
bool "Enable tracing for RCU - currently stats in debugfs"
select DEBUG_FS
default y
help
This option provides tracing in RCU which presents stats
in debugfs for debugging RCU implementation.
Say Y here if you want to enable RCU tracing
Say N if you are unsure.
Computing file changes ...
