The patch titled
percpu_counter: FBC_BATCH should be a variable
has been added to the -mm tree. Its filename is
percpu_counter-fbc_batch-should-be-a-variable.patch
Before you just go and hit "reply", please:
a) Consider who else should be cc'ed
b) Prefer to cc a suitable mailing list as well
c) Ideally: find the original patch on the mailing list and do a
reply-to-all to that, adding suitable additional cc's
*** Remember to use Documentation/SubmitChecklist when testing your code ***
See http://userweb.kernel.org/~akpm/stuff/added-to-mm.txt to find
out what to do about this
The current -mm tree may be found at http://userweb.kernel.org/~akpm/mmotm/
------------------------------------------------------
Subject: percpu_counter: FBC_BATCH should be a variable
From: Eric Dumazet <dada1@xxxxxxxxxxxxx>
For NR_CPUS >= 16 values, FBC_BATCH is 2*NR_CPUS
Considering more and more distros are using high NR_CPUS values, it makes
sense to use a more sensible value for FBC_BATCH, and get rid of NR_CPUS.
A sensible value is 2*num_online_cpus(), with a minimum value of 32 (This
minimum value helps branch prediction in __percpu_counter_add())
We already have a hotcpu notifier, so we can adjust FBC_BATCH dynamically.
We rename FBC_BATCH to percpu_counter_batch since its not a constant
anymore.
Signed-off-by: Eric Dumazet <dada1@xxxxxxxxxxxxx>
Acked-by: David S. Miller <davem@xxxxxxxxxxxxx>
Acked-by: Peter Zijlstra <a.p.zijlstra@xxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
fs/ext4/ext4.h | 6 +++---
fs/ext4/inode.c | 2 +-
include/linux/percpu_counter.h | 8 ++------
lib/percpu_counter.c | 18 ++++++++++++++----
4 files changed, 20 insertions(+), 14 deletions(-)
diff -puN fs/ext4/ext4.h~percpu_counter-fbc_batch-should-be-a-variable fs/ext4/ext4.h
--- a/fs/ext4/ext4.h~percpu_counter-fbc_batch-should-be-a-variable
+++ a/fs/ext4/ext4.h
@@ -1253,11 +1253,11 @@ do { \
} while (0)
#ifdef CONFIG_SMP
-/* Each CPU can accumulate FBC_BATCH blocks in their local
+/* Each CPU can accumulate percpu_counter_batch blocks in their local
* counters. So we need to make sure we have free blocks more
- * than FBC_BATCH * nr_cpu_ids. Also add a window of 4 times.
+ * than percpu_counter_batch * nr_cpu_ids. Also add a window of 4 times.
*/
-#define EXT4_FREEBLOCKS_WATERMARK (4 * (FBC_BATCH * nr_cpu_ids))
+#define EXT4_FREEBLOCKS_WATERMARK (4 * (percpu_counter_batch * nr_cpu_ids))
#else
#define EXT4_FREEBLOCKS_WATERMARK 0
#endif
diff -puN fs/ext4/inode.c~percpu_counter-fbc_batch-should-be-a-variable fs/ext4/inode.c
--- a/fs/ext4/inode.c~percpu_counter-fbc_batch-should-be-a-variable
+++ a/fs/ext4/inode.c
@@ -2533,7 +2533,7 @@ static int ext4_nonda_switch(struct supe
/*
* switch to non delalloc mode if we are running low
* on free block. The free block accounting via percpu
- * counters can get slightly wrong with FBC_BATCH getting
+ * counters can get slightly wrong with percpu_counter_batch getting
* accumulated on each CPU without updating global counters
* Delalloc need an accurate free block accounting. So switch
* to non delalloc when we are near to error range.
diff -puN include/linux/percpu_counter.h~percpu_counter-fbc_batch-should-be-a-variable include/linux/percpu_counter.h
--- a/include/linux/percpu_counter.h~percpu_counter-fbc_batch-should-be-a-variable
+++ a/include/linux/percpu_counter.h
@@ -24,11 +24,7 @@ struct percpu_counter {
s32 *counters;
};
-#if NR_CPUS >= 16
-#define FBC_BATCH (NR_CPUS*2)
-#else
-#define FBC_BATCH (NR_CPUS*4)
-#endif
+extern int percpu_counter_batch;
int percpu_counter_init(struct percpu_counter *fbc, s64 amount);
int percpu_counter_init_irq(struct percpu_counter *fbc, s64 amount);
@@ -39,7 +35,7 @@ s64 __percpu_counter_sum(struct percpu_c
static inline void percpu_counter_add(struct percpu_counter *fbc, s64 amount)
{
- __percpu_counter_add(fbc, amount, FBC_BATCH);
+ __percpu_counter_add(fbc, amount, percpu_counter_batch);
}
static inline s64 percpu_counter_sum_positive(struct percpu_counter *fbc)
diff -puN lib/percpu_counter.c~percpu_counter-fbc_batch-should-be-a-variable lib/percpu_counter.c
--- a/lib/percpu_counter.c~percpu_counter-fbc_batch-should-be-a-variable
+++ a/lib/percpu_counter.c
@@ -9,10 +9,8 @@
#include <linux/cpu.h>
#include <linux/module.h>
-#ifdef CONFIG_HOTPLUG_CPU
static LIST_HEAD(percpu_counters);
static DEFINE_MUTEX(percpu_counters_lock);
-#endif
void percpu_counter_set(struct percpu_counter *fbc, s64 amount)
{
@@ -111,13 +109,24 @@ void percpu_counter_destroy(struct percp
}
EXPORT_SYMBOL(percpu_counter_destroy);
-#ifdef CONFIG_HOTPLUG_CPU
+int percpu_counter_batch __read_mostly = 32;
+EXPORT_SYMBOL(percpu_counter_batch);
+
+static void compute_batch_value(void)
+{
+ int nr = num_online_cpus();
+
+ percpu_counter_batch = max(32, nr*2);
+}
+
static int __cpuinit percpu_counter_hotcpu_callback(struct notifier_block *nb,
unsigned long action, void *hcpu)
{
+#ifdef CONFIG_HOTPLUG_CPU
unsigned int cpu;
struct percpu_counter *fbc;
+ compute_batch_value();
if (action != CPU_DEAD)
return NOTIFY_OK;
@@ -134,13 +143,14 @@ static int __cpuinit percpu_counter_hotc
spin_unlock_irqrestore(&fbc->lock, flags);
}
mutex_unlock(&percpu_counters_lock);
+#endif
return NOTIFY_OK;
}
static int __init percpu_counter_startup(void)
{
+ compute_batch_value();
hotcpu_notifier(percpu_counter_hotcpu_callback, 0);
return 0;
}
module_init(percpu_counter_startup);
-#endif
_
Patches currently in -mm which might be from dada1@xxxxxxxxxxxxx are
origin.patch
percpu_counter-fix-cpu-unplug-race-in-percpu_counter_destroy.patch
revert-percpu-counter-clean-up-percpu_counter_sum_and_set.patch
revert-percpu_counter-new-function-percpu_counter_sum_and_set.patch
linux-next.patch
percpu_counter-fbc_batch-should-be-a-variable.patch
softirq-introduce-statistics-for-softirq.patch
proc-export-statistics-for-softirq-to-proc.patch
proc-update-document-for-proc-softirqs-and-proc-stat.patch
--
To unsubscribe from this list: send the line "unsubscribe mm-commits" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
