Prev: [PATCH v3 8/9] orinoco/wext.c: Remove local #define STD_IW_HANDLER
Next: [PATCH] Fix a typo in sound/pci/hda/hda_local.h
From: Mathieu Desnoyers on 19 Mar 2010 09:00
* Paul E. McKenney (paulmck(a)linux.vnet.ibm.com) wrote:
> On Fri, Mar 19, 2010 at 12:13:53PM +0800, Lai Jiangshan wrote:
> > CC: Christoph Lameter <cl(a)linux-foundation.org>
> > CC: Pekka Enberg <penberg(a)cs.helsinki.fi>
> >
> > I have written a patch like this before.
> >
> > but SLUB opposed me, so I did not post it.
> >
> > @ slub.c
> > static void free_slab(struct kmem_cache *s, struct page *page)
> > {
> > if (unlikely(s->flags & SLAB_DESTROY_BY_RCU)) {
> > /*
> > * RCU free overloads the RCU head over the LRU
> > */
> > struct rcu_head *head = (void *)&page->lru;
> >
> > call_rcu(head, rcu_free_slab);
> > } else
> > __free_slab(s, page);
> > }
> >
> > I think it is not your patch's fault, but the above code has
> > already existed.
> >
> > In your patch, struct rcu_head::debug will share the same memory of
> > the other field of struct page, (especially share the struct page::flags
> > which I encountered)
> >
> > I tested it very long time before, so I can't give you dmesg.
>
> Ouch!!!
>
> If this only causes a problem for SLUB, maybe a workaround is to make
> the config variable depend on !SLUB. But this would seem to increase
> the urgency of using the debug objects

Yep, that hurts. I think we'll have to move to debugobjects more quickly than
expected.

Thanks,

Matheiu


>
> Thanx, Paul
>
> > Lai
> >
> > Mathieu Desnoyers wrote:
> > > Poisoning the rcu_head callback list.
> > >
> > > Helps finding racy users of call_rcu(), which results in hangs because list
> > > entries are overwritten and/or skipped.
> > >
> > > A lot of non-initialized rcu list heads will be found with this option enabled.
> > > It is important to do not just blindly initialize them before each call_rcu, but
> > > rather to perform the initialization at the proper location, after the memory
> > > has been allocated, so we can effectively detect incorrect call_rcu users.
> > >
> > > This patch version does not use "debug objects", although it probably should.
> > > Some day I might find time to look into this transition, but the patch is usable
> > > as is.
> > >
> > > The weird #ifdef in the networking code comes from an agreement with Eric
> > > Dumazet about how to disable the build check for the networking code. For those
> > > who wonder about the existence of this build-time check: it generates a build
> > > error when the size of struct dst_entry grows (because it is very performance
> > > sensitive). So the agreement here has been to disable the check when the
> > > DEBUG_RCU_HEAD config option is active.
> > >
> > > Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers(a)efficios.com>
> > > CC: "Paul E. McKenney" <paulmck(a)linux.vnet.ibm.com>
> > > CC: mingo(a)elte.hu
> > > CC: akpm(a)linux-foundation.org
> > > CC: mingo(a)elte.hu
> > > CC: laijs(a)cn.fujitsu.com
> > > CC: dipankar(a)in.ibm.com
> > > CC: josh(a)joshtriplett.org
> > > CC: dvhltc(a)us.ibm.com
> > > CC: niv(a)us.ibm.com
> > > CC: tglx(a)linutronix.de
> > > CC: peterz(a)infradead.org
> > > CC: rostedt(a)goodmis.org
> > > CC: Valdis.Kletnieks(a)vt.edu
> > > CC: dhowells(a)redhat.com
> > > CC: eric.dumazet(a)gmail.com
> > > CC: Alexey Dobriyan <adobriyan(a)gmail.com>
> > > CC: Peter Zijlstra <a.p.zijlstra(a)chello.nl>
> > > ---
> > > include/linux/rcupdate.h | 13 ++++++++++++-
> > > include/net/dst.h | 2 ++
> > > kernel/rcutiny.c | 9 +++++++++
> > > kernel/rcutree.c | 10 ++++++++++
> > > lib/Kconfig.debug | 9 +++++++++
> > > 5 files changed, 42 insertions(+), 1 deletion(-)
> > >
> > > Index: linux-2.6-lttng/include/linux/rcupdate.h
> > > ===================================================================
> > > --- linux-2.6-lttng.orig/include/linux/rcupdate.h 2010-03-18 20:27:25.000000000 -0400
> > > +++ linux-2.6-lttng/include/linux/rcupdate.h 2010-03-18 20:30:54.000000000 -0400
> > > @@ -49,6 +49,9 @@
> > > struct rcu_head {
> > > struct rcu_head *next;
> > > void (*func)(struct rcu_head *head);
> > > +#ifdef CONFIG_DEBUG_RCU_HEAD
> > > + struct rcu_head *debug;
> > > +#endif
> > > };
> > >
> > > /* Exported common interfaces */
> > > @@ -71,11 +74,19 @@ extern void rcu_init(void);
> > > #error "Unknown RCU implementation specified to kernel configuration"
> > > #endif
> > >
> > > +#ifdef CONFIG_DEBUG_RCU_HEAD
> > > +#define RCU_HEAD_INIT { .next = NULL, .func = NULL, .debug = NULL }
> > > +#define INIT_RCU_HEAD(ptr) do { \
> > > + (ptr)->next = NULL; (ptr)->func = NULL; (ptr)->debug = NULL; \
> > > +} while (0)
> > > +#else
> > > #define RCU_HEAD_INIT { .next = NULL, .func = NULL }
> > > -#define RCU_HEAD(head) struct rcu_head head = RCU_HEAD_INIT
> > > #define INIT_RCU_HEAD(ptr) do { \
> > > (ptr)->next = NULL; (ptr)->func = NULL; \
> > > } while (0)
> > > +#endif
> > > +
> > > +#define RCU_HEAD(head) struct rcu_head head = RCU_HEAD_INIT
> > >
> > > #ifdef CONFIG_DEBUG_LOCK_ALLOC
> > > extern struct lockdep_map rcu_lock_map;
> > > Index: linux-2.6-lttng/kernel/rcutree.c
> > > ===================================================================
> > > --- linux-2.6-lttng.orig/kernel/rcutree.c 2010-03-18 20:27:25.000000000 -0400
> > > +++ linux-2.6-lttng/kernel/rcutree.c 2010-03-18 20:31:16.000000000 -0400
> > > @@ -39,6 +39,7 @@
> > > #include <asm/atomic.h>
> > > #include <linux/bitops.h>
> > > #include <linux/module.h>
> > > +#include <linux/poison.h>
> > > #include <linux/completion.h>
> > > #include <linux/moduleparam.h>
> > > #include <linux/percpu.h>
> > > @@ -1060,6 +1061,10 @@ static void rcu_do_batch(struct rcu_stat
> > > next = list->next;
> > > prefetch(next);
> > > trace_rcu_tree_callback(list);
> > > +#ifdef CONFIG_DEBUG_RCU_HEAD
> > > + WARN_ON_ONCE(list->debug != LIST_POISON1);
> > > + list->debug = NULL;
> > > +#endif
> > > list->func(list);
> > > list = next;
> > > if (++count >= rdp->blimit)
> > > @@ -1350,6 +1355,11 @@ __call_rcu(struct rcu_head *head, void (
> > > unsigned long flags;
> > > struct rcu_data *rdp;
> > >
> > > +#ifdef CONFIG_DEBUG_RCU_HEAD
> > > + WARN_ON_ONCE(head->debug);
> > > + head->debug = LIST_POISON1;
> > > +#endif
> > > +
> > > head->func = func;
> > > head->next = NULL;
> > >
> > > Index: linux-2.6-lttng/lib/Kconfig.debug
> > > ===================================================================
> > > --- linux-2.6-lttng.orig/lib/Kconfig.debug 2010-03-18 20:27:25.000000000 -0400
> > > +++ linux-2.6-lttng/lib/Kconfig.debug 2010-03-18 20:27:38.000000000 -0400
> > > @@ -661,6 +661,15 @@ config DEBUG_LIST
> > >
> > > If unsure, say N.
> > >
> > > +config DEBUG_RCU_HEAD
> > > + bool "Debug RCU callbacks"
> > > + depends on DEBUG_KERNEL
> > > + depends on TREE_RCU
> > > + help
> > > + Enable this to turn on debugging of RCU list heads (call_rcu() usage).
> > > + Seems to find problems more quickly with stress-tests in single-cpu
> > > + mode.
> > > +
> > > config DEBUG_SG
> > > bool "Debug SG table operations"
> > > depends on DEBUG_KERNEL
> > > Index: linux-2.6-lttng/include/net/dst.h
> > > ===================================================================
> > > --- linux-2.6-lttng.orig/include/net/dst.h 2010-03-18 20:27:25.000000000 -0400
> > > +++ linux-2.6-lttng/include/net/dst.h 2010-03-18 20:35:02.000000000 -0400
> > > @@ -159,7 +159,9 @@ static inline void dst_hold(struct dst_e
> > > * If your kernel compilation stops here, please check
> > > * __pad_to_align_refcnt declaration in struct dst_entry
> > > */
> > > +#ifndef CONFIG_DEBUG_RCU_HEAD
> > > BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
> > > +#endif
> > > atomic_inc(&dst->__refcnt);
> > > }
> > >
> > > Index: linux-2.6-lttng/kernel/rcutiny.c
> > > ===================================================================
> > > --- linux-2.6-lttng.orig/kernel/rcutiny.c 2010-03-18 20:35:14.000000000 -0400
> > > +++ linux-2.6-lttng/kernel/rcutiny.c 2010-03-18 20:39:12.000000000 -0400
> > > @@ -35,6 +35,7 @@
> > > #include <linux/init.h>
> > > #include <linux/time.h>
> > > #include <linux/cpu.h>
> > > +#include <linux/poison.h>
> > >
> > > /* Global control variables for rcupdate callback mechanism. */
> > > struct rcu_ctrlblk {
> > > @@ -163,6 +164,10 @@ static void __rcu_process_callbacks(stru
> > > while (list) {
> > > next = list->next;
> > > prefetch(next);
> > > +#ifdef CONFIG_DEBUG_RCU_HEAD
> > > + WARN_ON_ONCE(list->debug != LIST_POISON1);
> > > + list->debug = NULL;
> > > +#endif
> > > list->func(list);
> > > list = next;
> > > }
> > > @@ -210,6 +215,10 @@ static void __call_rcu(struct rcu_head *
> > > {
> > > unsigned long flags;
> > >
> > > +#ifdef CONFIG_DEBUG_RCU_HEAD
> > > + WARN_ON_ONCE(head->debug);
> > > + head->debug = LIST_POISON1;
> > > +#endif
> > > head->func = func;
> > > head->next = NULL;
> > >
> > >
> >

--
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Mathieu Desnoyers on 19 Mar 2010 10:00
* Paul E. McKenney (paulmck(a)linux.vnet.ibm.com) wrote:
> On Fri, Mar 19, 2010 at 12:13:53PM +0800, Lai Jiangshan wrote:
> > CC: Christoph Lameter <cl(a)linux-foundation.org>
> > CC: Pekka Enberg <penberg(a)cs.helsinki.fi>
> >
> > I have written a patch like this before.
> >
> > but SLUB opposed me, so I did not post it.
> >
> > @ slub.c
> > static void free_slab(struct kmem_cache *s, struct page *page)
> > {
> > if (unlikely(s->flags & SLAB_DESTROY_BY_RCU)) {
> > /*
> > * RCU free overloads the RCU head over the LRU
> > */
> > struct rcu_head *head = (void *)&page->lru;
> >
> > call_rcu(head, rcu_free_slab);
> > } else
> > __free_slab(s, page);
> > }
> >
> > I think it is not your patch's fault, but the above code has
> > already existed.
> >
> > In your patch, struct rcu_head::debug will share the same memory of
> > the other field of struct page, (especially share the struct page::flags
> > which I encountered)
> >
> > I tested it very long time before, so I can't give you dmesg.
>
> Ouch!!!
>
> If this only causes a problem for SLUB, maybe a workaround is to make
> the config variable depend on !SLUB. But this would seem to increase
> the urgency of using the debug objects

Hrm.. looking into the port to debugobjects, there are a few things that are not
precisely well set at the moment for it:

- We need to have two distinct active states, with a check that we move between
the two. This does not seem to be currently supported by debugobjects.
- RCU heads are never explicitely freed. So yeah, that means we will have to
modify a lot of callers.

Thanks,

Mathieu


>
> Thanx, Paul
>
> > Lai
> >
> > Mathieu Desnoyers wrote:
> > > Poisoning the rcu_head callback list.
> > >
> > > Helps finding racy users of call_rcu(), which results in hangs because list
> > > entries are overwritten and/or skipped.
> > >
> > > A lot of non-initialized rcu list heads will be found with this option enabled.
> > > It is important to do not just blindly initialize them before each call_rcu, but
> > > rather to perform the initialization at the proper location, after the memory
> > > has been allocated, so we can effectively detect incorrect call_rcu users.
> > >
> > > This patch version does not use "debug objects", although it probably should.
> > > Some day I might find time to look into this transition, but the patch is usable
> > > as is.
> > >
> > > The weird #ifdef in the networking code comes from an agreement with Eric
> > > Dumazet about how to disable the build check for the networking code. For those
> > > who wonder about the existence of this build-time check: it generates a build
> > > error when the size of struct dst_entry grows (because it is very performance
> > > sensitive). So the agreement here has been to disable the check when the
> > > DEBUG_RCU_HEAD config option is active.
> > >
> > > Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers(a)efficios.com>
> > > CC: "Paul E. McKenney" <paulmck(a)linux.vnet.ibm.com>
> > > CC: mingo(a)elte.hu
> > > CC: akpm(a)linux-foundation.org
> > > CC: mingo(a)elte.hu
> > > CC: laijs(a)cn.fujitsu.com
> > > CC: dipankar(a)in.ibm.com
> > > CC: josh(a)joshtriplett.org
> > > CC: dvhltc(a)us.ibm.com
> > > CC: niv(a)us.ibm.com
> > > CC: tglx(a)linutronix.de
> > > CC: peterz(a)infradead.org
> > > CC: rostedt(a)goodmis.org
> > > CC: Valdis.Kletnieks(a)vt.edu
> > > CC: dhowells(a)redhat.com
> > > CC: eric.dumazet(a)gmail.com
> > > CC: Alexey Dobriyan <adobriyan(a)gmail.com>
> > > CC: Peter Zijlstra <a.p.zijlstra(a)chello.nl>
> > > ---
> > > include/linux/rcupdate.h | 13 ++++++++++++-
> > > include/net/dst.h | 2 ++
> > > kernel/rcutiny.c | 9 +++++++++
> > > kernel/rcutree.c | 10 ++++++++++
> > > lib/Kconfig.debug | 9 +++++++++
> > > 5 files changed, 42 insertions(+), 1 deletion(-)
> > >
> > > Index: linux-2.6-lttng/include/linux/rcupdate.h
> > > ===================================================================
> > > --- linux-2.6-lttng.orig/include/linux/rcupdate.h 2010-03-18 20:27:25.000000000 -0400
> > > +++ linux-2.6-lttng/include/linux/rcupdate.h 2010-03-18 20:30:54.000000000 -0400
> > > @@ -49,6 +49,9 @@
> > > struct rcu_head {
> > > struct rcu_head *next;
> > > void (*func)(struct rcu_head *head);
> > > +#ifdef CONFIG_DEBUG_RCU_HEAD
> > > + struct rcu_head *debug;
> > > +#endif
> > > };
> > >
> > > /* Exported common interfaces */
> > > @@ -71,11 +74,19 @@ extern void rcu_init(void);
> > > #error "Unknown RCU implementation specified to kernel configuration"
> > > #endif
> > >
> > > +#ifdef CONFIG_DEBUG_RCU_HEAD
> > > +#define RCU_HEAD_INIT { .next = NULL, .func = NULL, .debug = NULL }
> > > +#define INIT_RCU_HEAD(ptr) do { \
> > > + (ptr)->next = NULL; (ptr)->func = NULL; (ptr)->debug = NULL; \
> > > +} while (0)
> > > +#else
> > > #define RCU_HEAD_INIT { .next = NULL, .func = NULL }
> > > -#define RCU_HEAD(head) struct rcu_head head = RCU_HEAD_INIT
> > > #define INIT_RCU_HEAD(ptr) do { \
> > > (ptr)->next = NULL; (ptr)->func = NULL; \
> > > } while (0)
> > > +#endif
> > > +
> > > +#define RCU_HEAD(head) struct rcu_head head = RCU_HEAD_INIT
> > >
> > > #ifdef CONFIG_DEBUG_LOCK_ALLOC
> > > extern struct lockdep_map rcu_lock_map;
> > > Index: linux-2.6-lttng/kernel/rcutree.c
> > > ===================================================================
> > > --- linux-2.6-lttng.orig/kernel/rcutree.c 2010-03-18 20:27:25.000000000 -0400
> > > +++ linux-2.6-lttng/kernel/rcutree.c 2010-03-18 20:31:16.000000000 -0400
> > > @@ -39,6 +39,7 @@
> > > #include <asm/atomic.h>
> > > #include <linux/bitops.h>
> > > #include <linux/module.h>
> > > +#include <linux/poison.h>
> > > #include <linux/completion.h>
> > > #include <linux/moduleparam.h>
> > > #include <linux/percpu.h>
> > > @@ -1060,6 +1061,10 @@ static void rcu_do_batch(struct rcu_stat
> > > next = list->next;
> > > prefetch(next);
> > > trace_rcu_tree_callback(list);
> > > +#ifdef CONFIG_DEBUG_RCU_HEAD
> > > + WARN_ON_ONCE(list->debug != LIST_POISON1);
> > > + list->debug = NULL;
> > > +#endif
> > > list->func(list);
> > > list = next;
> > > if (++count >= rdp->blimit)
> > > @@ -1350,6 +1355,11 @@ __call_rcu(struct rcu_head *head, void (
> > > unsigned long flags;
> > > struct rcu_data *rdp;
> > >
> > > +#ifdef CONFIG_DEBUG_RCU_HEAD
> > > + WARN_ON_ONCE(head->debug);
> > > + head->debug = LIST_POISON1;
> > > +#endif
> > > +
> > > head->func = func;
> > > head->next = NULL;
> > >
> > > Index: linux-2.6-lttng/lib/Kconfig.debug
> > > ===================================================================
> > > --- linux-2.6-lttng.orig/lib/Kconfig.debug 2010-03-18 20:27:25.000000000 -0400
> > > +++ linux-2.6-lttng/lib/Kconfig.debug 2010-03-18 20:27:38.000000000 -0400
> > > @@ -661,6 +661,15 @@ config DEBUG_LIST
> > >
> > > If unsure, say N.
> > >
> > > +config DEBUG_RCU_HEAD
> > > + bool "Debug RCU callbacks"
> > > + depends on DEBUG_KERNEL
> > > + depends on TREE_RCU
> > > + help
> > > + Enable this to turn on debugging of RCU list heads (call_rcu() usage).
> > > + Seems to find problems more quickly with stress-tests in single-cpu
> > > + mode.
> > > +
> > > config DEBUG_SG
> > > bool "Debug SG table operations"
> > > depends on DEBUG_KERNEL
> > > Index: linux-2.6-lttng/include/net/dst.h
> > > ===================================================================
> > > --- linux-2.6-lttng.orig/include/net/dst.h 2010-03-18 20:27:25.000000000 -0400
> > > +++ linux-2.6-lttng/include/net/dst.h 2010-03-18 20:35:02.000000000 -0400
> > > @@ -159,7 +159,9 @@ static inline void dst_hold(struct dst_e
> > > * If your kernel compilation stops here, please check
> > > * __pad_to_align_refcnt declaration in struct dst_entry
> > > */
> > > +#ifndef CONFIG_DEBUG_RCU_HEAD
> > > BUILD_BUG_ON(offsetof(struct dst_entry, __refcnt) & 63);
> > > +#endif
> > > atomic_inc(&dst->__refcnt);
> > > }
> > >
> > > Index: linux-2.6-lttng/kernel/rcutiny.c
> > > ===================================================================
> > > --- linux-2.6-lttng.orig/kernel/rcutiny.c 2010-03-18 20:35:14.000000000 -0400
> > > +++ linux-2.6-lttng/kernel/rcutiny.c 2010-03-18 20:39:12.000000000 -0400
> > > @@ -35,6 +35,7 @@
> > > #include <linux/init.h>
> > > #include <linux/time.h>
> > > #include <linux/cpu.h>
> > > +#include <linux/poison.h>
> > >
> > > /* Global control variables for rcupdate callback mechanism. */
> > > struct rcu_ctrlblk {
> > > @@ -163,6 +164,10 @@ static void __rcu_process_callbacks(stru
> > > while (list) {
> > > next = list->next;
> > > prefetch(next);
> > > +#ifdef CONFIG_DEBUG_RCU_HEAD
> > > + WARN_ON_ONCE(list->debug != LIST_POISON1);
> > > + list->debug = NULL;
> > > +#endif
> > > list->func(list);
> > > list = next;
> > > }
> > > @@ -210,6 +215,10 @@ static void __call_rcu(struct rcu_head *
> > > {
> > > unsigned long flags;
> > >
> > > +#ifdef CONFIG_DEBUG_RCU_HEAD
> > > + WARN_ON_ONCE(head->debug);
> > > + head->debug = LIST_POISON1;
> > > +#endif
> > > head->func = func;
> > > head->next = NULL;
> > >
> > >
> >

--
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Paul E. McKenney on 19 Mar 2010 10:50
On Fri, Mar 19, 2010 at 07:34:07AM -0500, Christoph Lameter wrote:
> On Fri, 19 Mar 2010, Lai Jiangshan wrote:
>
> > but SLUB opposed me, so I did not post it.
> >
> > @ slub.c
> > static void free_slab(struct kmem_cache *s, struct page *page)
> > {
> > if (unlikely(s->flags & SLAB_DESTROY_BY_RCU)) {
> > /*
> > * RCU free overloads the RCU head over the LRU
> > */
> > struct rcu_head *head = (void *)&page->lru;
> >
> > call_rcu(head, rcu_free_slab);
> > } else
> > __free_slab(s, page);
> > }
> >
> > I think it is not your patch's fault, but the above code has
> > already existed.
> >
> > In your patch, struct rcu_head::debug will share the same memory of
> > the other field of struct page, (especially share the struct page::flags
> > which I encountered)
>
> Right. I like the smallness of the RCU structure but if you must
> increase its size then we could:
>
> A) put the rcu structure at the end of each page like SLOB
> but that will take away memory that could be used by objects.
>
> B) Move the rcu object up by a word so that it overloads the index field
> in the page struct. That would give another exact fit but will cause
> trouble again if you bloat the structure some more.

The size increase is strictly under a debug option -- the non-debug
size remains the same as always. And I expect that in the longer term,
we will be able to shrink the size back down even with this debug option
available.

One question: do any of the other allocators care about the size of the
struct rcu_head? If not, a temporary work-around would be to simply
prohibit use of this debug option when running with SLUB.

Thanx, Paul
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Mathieu Desnoyers on 19 Mar 2010 13:00
* Paul E. McKenney (paulmck(a)linux.vnet.ibm.com) wrote:
> On Fri, Mar 19, 2010 at 07:34:07AM -0500, Christoph Lameter wrote:
> > On Fri, 19 Mar 2010, Lai Jiangshan wrote:
> >
> > > but SLUB opposed me, so I did not post it.
> > >
> > > @ slub.c
> > > static void free_slab(struct kmem_cache *s, struct page *page)
> > > {
> > > if (unlikely(s->flags & SLAB_DESTROY_BY_RCU)) {
> > > /*
> > > * RCU free overloads the RCU head over the LRU
> > > */
> > > struct rcu_head *head = (void *)&page->lru;
> > >
> > > call_rcu(head, rcu_free_slab);
> > > } else
> > > __free_slab(s, page);
> > > }
> > >
> > > I think it is not your patch's fault, but the above code has
> > > already existed.
> > >
> > > In your patch, struct rcu_head::debug will share the same memory of
> > > the other field of struct page, (especially share the struct page::flags
> > > which I encountered)
> >
> > Right. I like the smallness of the RCU structure but if you must
> > increase its size then we could:
> >
> > A) put the rcu structure at the end of each page like SLOB
> > but that will take away memory that could be used by objects.
> >
> > B) Move the rcu object up by a word so that it overloads the index field
> > in the page struct. That would give another exact fit but will cause
> > trouble again if you bloat the structure some more.
>
> The size increase is strictly under a debug option -- the non-debug
> size remains the same as always. And I expect that in the longer term,
> we will be able to shrink the size back down even with this debug option
> available.
>
> One question: do any of the other allocators care about the size of the
> struct rcu_head? If not, a temporary work-around would be to simply
> prohibit use of this debug option when running with SLUB.
>
> Thanx, Paul

Don't worry about the struct rcu_head size too much. I'm currently working on a
debugobjects implementation. I should have something by the end of the day.

Thanks,

Mathieu


--
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
From: Paul E. McKenney on 19 Mar 2010 13:20
On Fri, Mar 19, 2010 at 12:56:38PM -0400, Mathieu Desnoyers wrote:
> * Paul E. McKenney (paulmck(a)linux.vnet.ibm.com) wrote:
> > On Fri, Mar 19, 2010 at 07:34:07AM -0500, Christoph Lameter wrote:
> > > On Fri, 19 Mar 2010, Lai Jiangshan wrote:
> > >
> > > > but SLUB opposed me, so I did not post it.
> > > >
> > > > @ slub.c
> > > > static void free_slab(struct kmem_cache *s, struct page *page)
> > > > {
> > > > if (unlikely(s->flags & SLAB_DESTROY_BY_RCU)) {
> > > > /*
> > > > * RCU free overloads the RCU head over the LRU
> > > > */
> > > > struct rcu_head *head = (void *)&page->lru;
> > > >
> > > > call_rcu(head, rcu_free_slab);
> > > > } else
> > > > __free_slab(s, page);
> > > > }
> > > >
> > > > I think it is not your patch's fault, but the above code has
> > > > already existed.
> > > >
> > > > In your patch, struct rcu_head::debug will share the same memory of
> > > > the other field of struct page, (especially share the struct page::flags
> > > > which I encountered)
> > >
> > > Right. I like the smallness of the RCU structure but if you must
> > > increase its size then we could:
> > >
> > > A) put the rcu structure at the end of each page like SLOB
> > > but that will take away memory that could be used by objects.
> > >
> > > B) Move the rcu object up by a word so that it overloads the index field
> > > in the page struct. That would give another exact fit but will cause
> > > trouble again if you bloat the structure some more.
> >
> > The size increase is strictly under a debug option -- the non-debug
> > size remains the same as always. And I expect that in the longer term,
> > we will be able to shrink the size back down even with this debug option
> > available.
> >
> > One question: do any of the other allocators care about the size of the
> > struct rcu_head? If not, a temporary work-around would be to simply
> > prohibit use of this debug option when running with SLUB.
> >
> > Thanx, Paul
>
> Don't worry about the struct rcu_head size too much. I'm currently working on a
> debugobjects implementation. I should have something by the end of the day.

Even better!!! Thank you, Mathieu!!!

Thanx, Paul
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo(a)vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
First  |  Prev  |  Next  |  Last
Pages: 1 2 3
Prev: [PATCH v3 8/9] orinoco/wext.c: Remove local #define STD_IW_HANDLER
Next: [PATCH] Fix a typo in sound/pci/hda/hda_local.h