2008-07-28 11:06:00 +00:00
|
|
|
/*
|
|
|
|
* Functions related to softirq rq completions
|
|
|
|
*/
|
|
|
|
#include <linux/kernel.h>
|
|
|
|
#include <linux/module.h>
|
|
|
|
#include <linux/init.h>
|
|
|
|
#include <linux/bio.h>
|
|
|
|
#include <linux/blkdev.h>
|
|
|
|
#include <linux/interrupt.h>
|
|
|
|
#include <linux/cpu.h>
|
|
|
|
|
|
|
|
#include "blk.h"
|
|
|
|
|
|
|
|
static DEFINE_PER_CPU(struct list_head, blk_cpu_done);
|
|
|
|
|
2008-09-13 18:26:01 +00:00
|
|
|
/*
|
|
|
|
* Softirq action handler - move entries to local list and loop over them
|
|
|
|
* while passing them to the queue registered handler.
|
|
|
|
*/
|
|
|
|
static void blk_done_softirq(struct softirq_action *h)
|
|
|
|
{
|
|
|
|
struct list_head *cpu_list, local_list;
|
|
|
|
|
|
|
|
local_irq_disable();
|
|
|
|
cpu_list = &__get_cpu_var(blk_cpu_done);
|
|
|
|
list_replace_init(cpu_list, &local_list);
|
|
|
|
local_irq_enable();
|
|
|
|
|
|
|
|
while (!list_empty(&local_list)) {
|
|
|
|
struct request *rq;
|
|
|
|
|
|
|
|
rq = list_entry(local_list.next, struct request, csd.list);
|
|
|
|
list_del_init(&rq->csd.list);
|
|
|
|
rq->q->softirq_done_fn(rq);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
#if defined(CONFIG_SMP) && defined(CONFIG_USE_GENERIC_SMP_HELPERS)
|
|
|
|
static void trigger_softirq(void *data)
|
|
|
|
{
|
|
|
|
struct request *rq = data;
|
|
|
|
unsigned long flags;
|
|
|
|
struct list_head *list;
|
|
|
|
|
|
|
|
local_irq_save(flags);
|
|
|
|
list = &__get_cpu_var(blk_cpu_done);
|
|
|
|
list_add_tail(&rq->csd.list, list);
|
|
|
|
|
|
|
|
if (list->next == &rq->csd.list)
|
|
|
|
raise_softirq_irqoff(BLOCK_SOFTIRQ);
|
|
|
|
|
|
|
|
local_irq_restore(flags);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Setup and invoke a run of 'trigger_softirq' on the given cpu.
|
|
|
|
*/
|
|
|
|
static int raise_blk_irq(int cpu, struct request *rq)
|
|
|
|
{
|
|
|
|
if (cpu_online(cpu)) {
|
|
|
|
struct call_single_data *data = &rq->csd;
|
|
|
|
|
|
|
|
data->func = trigger_softirq;
|
|
|
|
data->info = rq;
|
|
|
|
data->flags = 0;
|
|
|
|
|
2009-02-25 12:59:48 +00:00
|
|
|
__smp_call_function_single(cpu, data, 0);
|
2008-09-13 18:26:01 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
#else /* CONFIG_SMP && CONFIG_USE_GENERIC_SMP_HELPERS */
|
|
|
|
static int raise_blk_irq(int cpu, struct request *rq)
|
|
|
|
{
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2008-07-28 11:06:00 +00:00
|
|
|
static int __cpuinit blk_cpu_notify(struct notifier_block *self,
|
|
|
|
unsigned long action, void *hcpu)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* If a CPU goes away, splice its entries to the current CPU
|
|
|
|
* and trigger a run of the softirq
|
|
|
|
*/
|
|
|
|
if (action == CPU_DEAD || action == CPU_DEAD_FROZEN) {
|
|
|
|
int cpu = (unsigned long) hcpu;
|
|
|
|
|
|
|
|
local_irq_disable();
|
|
|
|
list_splice_init(&per_cpu(blk_cpu_done, cpu),
|
|
|
|
&__get_cpu_var(blk_cpu_done));
|
|
|
|
raise_softirq_irqoff(BLOCK_SOFTIRQ);
|
|
|
|
local_irq_enable();
|
|
|
|
}
|
|
|
|
|
|
|
|
return NOTIFY_OK;
|
|
|
|
}
|
|
|
|
|
2008-09-13 18:26:01 +00:00
|
|
|
static struct notifier_block __cpuinitdata blk_cpu_notifier = {
|
2008-07-28 11:06:00 +00:00
|
|
|
.notifier_call = blk_cpu_notify,
|
|
|
|
};
|
|
|
|
|
2008-09-14 12:55:09 +00:00
|
|
|
void __blk_complete_request(struct request *req)
|
2008-07-28 11:06:00 +00:00
|
|
|
{
|
block: improve rq_affinity placement
This patch reverts commit 35ae66e0a09ab70ed(block: Make rq_affinity = 1
work as expected). The purpose is to avoid an unnecessary IPI.
Let's take an example. My test box has cpu 0-7, one socket. Say request is
added from CPU 1, blk_complete_request() occurs at CPU 7. Without the reverted
patch, softirq will be done at CPU 7. With it, an IPI will be directed to CPU
0, and softirq will be done at CPU 0. In this case, doing softirq at CPU 0 and
CPU 7 have no difference from cache sharing point view and we can avoid an
ipi if doing it in CPU 7.
An immediate concern is this is just like QUEUE_FLAG_SAME_FORCE, but actually
not. blk_complete_request() is running in interrupt handler, and currently
I/O controller doesn't support multiple interrupts (I checked several LSI
cards and AHCI), so only one CPU can run blk_complete_request(). This is
still quite different as QUEUE_FLAG_SAME_FORCE.
Since only one CPU runs softirq, the only difference with below patch is
softirq not always runs at the first CPU of a group.
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2011-08-11 08:39:04 +00:00
|
|
|
int ccpu, cpu, group_cpu = NR_CPUS;
|
2008-09-13 18:26:01 +00:00
|
|
|
struct request_queue *q = req->q;
|
2008-07-28 11:06:00 +00:00
|
|
|
unsigned long flags;
|
|
|
|
|
2008-09-13 18:26:01 +00:00
|
|
|
BUG_ON(!q->softirq_done_fn);
|
2008-07-28 11:06:00 +00:00
|
|
|
|
|
|
|
local_irq_save(flags);
|
2008-09-13 18:26:01 +00:00
|
|
|
cpu = smp_processor_id();
|
2008-07-28 11:06:00 +00:00
|
|
|
|
2008-09-13 18:26:01 +00:00
|
|
|
/*
|
|
|
|
* Select completion CPU
|
|
|
|
*/
|
2011-09-14 07:31:01 +00:00
|
|
|
if (req->cpu != -1) {
|
2008-09-13 18:26:01 +00:00
|
|
|
ccpu = req->cpu;
|
block: improve rq_affinity placement
This patch reverts commit 35ae66e0a09ab70ed(block: Make rq_affinity = 1
work as expected). The purpose is to avoid an unnecessary IPI.
Let's take an example. My test box has cpu 0-7, one socket. Say request is
added from CPU 1, blk_complete_request() occurs at CPU 7. Without the reverted
patch, softirq will be done at CPU 7. With it, an IPI will be directed to CPU
0, and softirq will be done at CPU 0. In this case, doing softirq at CPU 0 and
CPU 7 have no difference from cache sharing point view and we can avoid an
ipi if doing it in CPU 7.
An immediate concern is this is just like QUEUE_FLAG_SAME_FORCE, but actually
not. blk_complete_request() is running in interrupt handler, and currently
I/O controller doesn't support multiple interrupts (I checked several LSI
cards and AHCI), so only one CPU can run blk_complete_request(). This is
still quite different as QUEUE_FLAG_SAME_FORCE.
Since only one CPU runs softirq, the only difference with below patch is
softirq not always runs at the first CPU of a group.
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2011-08-11 08:39:04 +00:00
|
|
|
if (!test_bit(QUEUE_FLAG_SAME_FORCE, &q->queue_flags)) {
|
2011-07-23 18:44:25 +00:00
|
|
|
ccpu = blk_cpu_to_group(ccpu);
|
block: improve rq_affinity placement
This patch reverts commit 35ae66e0a09ab70ed(block: Make rq_affinity = 1
work as expected). The purpose is to avoid an unnecessary IPI.
Let's take an example. My test box has cpu 0-7, one socket. Say request is
added from CPU 1, blk_complete_request() occurs at CPU 7. Without the reverted
patch, softirq will be done at CPU 7. With it, an IPI will be directed to CPU
0, and softirq will be done at CPU 0. In this case, doing softirq at CPU 0 and
CPU 7 have no difference from cache sharing point view and we can avoid an
ipi if doing it in CPU 7.
An immediate concern is this is just like QUEUE_FLAG_SAME_FORCE, but actually
not. blk_complete_request() is running in interrupt handler, and currently
I/O controller doesn't support multiple interrupts (I checked several LSI
cards and AHCI), so only one CPU can run blk_complete_request(). This is
still quite different as QUEUE_FLAG_SAME_FORCE.
Since only one CPU runs softirq, the only difference with below patch is
softirq not always runs at the first CPU of a group.
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2011-08-11 08:39:04 +00:00
|
|
|
group_cpu = blk_cpu_to_group(cpu);
|
|
|
|
}
|
2011-07-23 18:44:25 +00:00
|
|
|
} else
|
2008-09-13 18:26:01 +00:00
|
|
|
ccpu = cpu;
|
|
|
|
|
block: improve rq_affinity placement
This patch reverts commit 35ae66e0a09ab70ed(block: Make rq_affinity = 1
work as expected). The purpose is to avoid an unnecessary IPI.
Let's take an example. My test box has cpu 0-7, one socket. Say request is
added from CPU 1, blk_complete_request() occurs at CPU 7. Without the reverted
patch, softirq will be done at CPU 7. With it, an IPI will be directed to CPU
0, and softirq will be done at CPU 0. In this case, doing softirq at CPU 0 and
CPU 7 have no difference from cache sharing point view and we can avoid an
ipi if doing it in CPU 7.
An immediate concern is this is just like QUEUE_FLAG_SAME_FORCE, but actually
not. blk_complete_request() is running in interrupt handler, and currently
I/O controller doesn't support multiple interrupts (I checked several LSI
cards and AHCI), so only one CPU can run blk_complete_request(). This is
still quite different as QUEUE_FLAG_SAME_FORCE.
Since only one CPU runs softirq, the only difference with below patch is
softirq not always runs at the first CPU of a group.
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
2011-08-11 08:39:04 +00:00
|
|
|
/*
|
|
|
|
* If current CPU and requested CPU are in the same group, running
|
|
|
|
* softirq in current CPU. One might concern this is just like
|
|
|
|
* QUEUE_FLAG_SAME_FORCE, but actually not. blk_complete_request() is
|
|
|
|
* running in interrupt handler, and currently I/O controller doesn't
|
|
|
|
* support multiple interrupts, so current CPU is unique actually. This
|
|
|
|
* avoids IPI sending from current CPU to the first CPU of a group.
|
|
|
|
*/
|
|
|
|
if (ccpu == cpu || ccpu == group_cpu) {
|
2008-09-13 18:26:01 +00:00
|
|
|
struct list_head *list;
|
|
|
|
do_local:
|
|
|
|
list = &__get_cpu_var(blk_cpu_done);
|
|
|
|
list_add_tail(&req->csd.list, list);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* if the list only contains our just added request,
|
|
|
|
* signal a raise of the softirq. If there are already
|
|
|
|
* entries there, someone already raised the irq but it
|
|
|
|
* hasn't run yet.
|
|
|
|
*/
|
|
|
|
if (list->next == &req->csd.list)
|
|
|
|
raise_softirq_irqoff(BLOCK_SOFTIRQ);
|
|
|
|
} else if (raise_blk_irq(ccpu, req))
|
|
|
|
goto do_local;
|
2008-07-28 11:06:00 +00:00
|
|
|
|
|
|
|
local_irq_restore(flags);
|
|
|
|
}
|
2008-09-14 12:55:09 +00:00
|
|
|
|
|
|
|
/**
|
|
|
|
* blk_complete_request - end I/O on a request
|
|
|
|
* @req: the request being processed
|
|
|
|
*
|
|
|
|
* Description:
|
|
|
|
* Ends all I/O on a request. It does not handle partial completions,
|
|
|
|
* unless the driver actually implements this in its completion callback
|
|
|
|
* through requeueing. The actual completion happens out-of-order,
|
|
|
|
* through a softirq handler. The user must have registered a completion
|
|
|
|
* callback through blk_queue_softirq_done().
|
|
|
|
**/
|
|
|
|
void blk_complete_request(struct request *req)
|
|
|
|
{
|
2008-09-14 12:56:33 +00:00
|
|
|
if (unlikely(blk_should_fake_timeout(req->q)))
|
|
|
|
return;
|
2008-09-14 12:55:09 +00:00
|
|
|
if (!blk_mark_rq_complete(req))
|
|
|
|
__blk_complete_request(req);
|
|
|
|
}
|
2008-07-28 11:06:00 +00:00
|
|
|
EXPORT_SYMBOL(blk_complete_request);
|
|
|
|
|
2008-12-10 14:47:33 +00:00
|
|
|
static __init int blk_softirq_init(void)
|
2008-07-28 11:06:00 +00:00
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for_each_possible_cpu(i)
|
|
|
|
INIT_LIST_HEAD(&per_cpu(blk_cpu_done, i));
|
|
|
|
|
|
|
|
open_softirq(BLOCK_SOFTIRQ, blk_done_softirq);
|
|
|
|
register_hotcpu_notifier(&blk_cpu_notifier);
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
subsys_initcall(blk_softirq_init);
|