Hi,
The following patch adds ioprio to blktrace.
Explanation:
As you know, I/O schedulers such as cfq use io_contexts of current
tasks to schedule block I/O. But recently, some are suggesting to
use more appropriate io_contexts obtained by, for example, doing
io-tracking, making struct bio have io_context member, etc. I thought
adding ioprio to blktrace output might be helpful.
Implementation:
1. I used req_get_ioprio() and [new]bio_get_ioprio() to get ioprio.
Currently bio_get_ioprio just returns current->io_context->ioprio.
Appreciate any comments:
Does this kind of information helpful?
Thanks,
Takuya Yoshikawa
Example:
8,16 0 1 0.000000000 2664 Q R ioprio=16386 81920 + 8
[tiotest]
8,16 0 2 0.000010493 2664 G R ioprio=16386 81920 + 8
[tiotest]
8,16 0 3 0.000013217 2664 P N ioprio= 0 [tiotest]
8,16 0 4 0.000014164 2664 I R ioprio= 0 81920 + 8
[tiotest]
8,16 0 5 0.000019738 2664 U N ioprio= 0 [tiotest] 2
8,16 0 6 0.000026737 2664 D R ioprio= 0 81920 + 8
[tiotest]
8,16 0 7 0.005455128 2670 Q R ioprio=16389 327680 + 8
[tiotest]
8,16 0 8 0.005459912 2670 G R ioprio=16389 327680 + 8
[tiotest]
8,16 0 9 0.005460425 2670 P N ioprio= 0 [tiotest]
8,16 0 10 0.005460703 2670 I R ioprio= 0 327680 + 8
[tiotest]
8,16 0 11 0.005462495 2670 U N ioprio= 0 [tiotest] 7
==
Signed-off-by: Takuya Yoshikawa <yoshikawa.takuya at oss.ntt.co.jp>
---
diff -uprN linux-2.6.27-rc5-mm1/block/blktrace.c
linux-2.6.27-rc5-mm1-blktrace/block/blktrace.c
--- linux-2.6.27-rc5-mm1/block/blktrace.c 2008-09-10 19:27:44.000000000 +0900
+++ linux-2.6.27-rc5-mm1-blktrace/block/blktrace.c 2008-09-11 00:44:52.000000000
+0900
@@ -120,7 +120,8 @@ static u32 ddir_act[2] __read_mostly = {
* blk_io_trace structure and places it in a per-cpu subbuffer.
*/
void __blk_add_trace(struct blk_trace *bt, sector_t sector, int bytes,
- int rw, u32 what, int error, int pdu_len, void *pdu_data)
+ int rw, u32 what, int error, unsigned short ioprio,
+ int pdu_len, void *pdu_data)
{
struct task_struct *tsk = current;
struct blk_io_trace *t;
@@ -168,6 +169,7 @@ void __blk_add_trace(struct blk_trace *b
t->device = bt->dev;
t->cpu = cpu;
t->error = error;
+ t->ioprio = ioprio;
t->pdu_len = pdu_len;
if (pdu_len)
diff -uprN linux-2.6.27-rc5-mm1/include/linux/bio.h
linux-2.6.27-rc5-mm1-blktrace/include/linux/bio.h
--- linux-2.6.27-rc5-mm1/include/linux/bio.h 2008-09-10 19:27:46.000000000 +0900
+++ linux-2.6.27-rc5-mm1-blktrace/include/linux/bio.h 2008-09-11
00:53:15.000000000 +0900
@@ -186,6 +186,18 @@ static inline void *bio_data(struct bio
}
/*
+ * TODO: replace this with io-tracking version
+ */
+static inline unsigned short bio_get_ioprio(struct bio *bio)
+{
+ struct task_struct *tsk = current;
+ if (tsk->io_context)
+ return tsk->io_context->ioprio;
+
+ return 0;
+}
+
+/*
* will die
*/
#define bio_to_phys(bio) (page_to_phys(bio_page((bio))) + (unsigned long)
bio_offset((bio)))
diff -uprN linux-2.6.27-rc5-mm1/include/linux/blktrace_api.h
linux-2.6.27-rc5-mm1-blktrace/include/linux/blktrace_api.h
--- linux-2.6.27-rc5-mm1/include/linux/blktrace_api.h 2008-09-10
19:27:46.000000000 +0900
+++ linux-2.6.27-rc5-mm1-blktrace/include/linux/blktrace_api.h 2008-09-11
00:54:35.000000000 +0900
@@ -103,6 +103,7 @@ struct blk_io_trace {
u32 cpu; /* on what cpu did it happen */
u16 error; /* completion error */
u16 pdu_len; /* length of data after this trace */
+ u16 ioprio; /* priority of this io */
};
/*
@@ -153,7 +154,7 @@ struct blk_user_trace_setup {
#if defined(CONFIG_BLK_DEV_IO_TRACE)
extern int blk_trace_ioctl(struct block_device *, unsigned, char __user *);
extern void blk_trace_shutdown(struct request_queue *);
-extern void __blk_add_trace(struct blk_trace *, sector_t, int, int, u32, int,
int, void *);
+extern void __blk_add_trace(struct blk_trace *, sector_t, int, int, u32, int,
unsigned short, int, void *);
extern int do_blk_trace_setup(struct request_queue *q,
char *name, dev_t dev, struct blk_user_trace_setup *buts);
extern void __trace_note_message(struct blk_trace *, const char *fmt, ...);
@@ -194,6 +195,7 @@ static inline void blk_add_trace_rq(stru
{
struct blk_trace *bt = q->blk_trace;
int rw = rq->cmd_flags & 0x03;
+ unsigned short ioprio = req_get_ioprio(rq);
if (likely(!bt))
return;
@@ -203,10 +205,10 @@ static inline void blk_add_trace_rq(stru
if (blk_pc_request(rq)) {
what |= BLK_TC_ACT(BLK_TC_PC);
- __blk_add_trace(bt, 0, rq->data_len, rw, what, rq->errors,
sizeof(rq->cmd), rq->cmd);
+ __blk_add_trace(bt, 0, rq->data_len, rw, what, rq->errors, ioprio,
sizeof(rq->cmd), rq->cmd);
} else {
what |= BLK_TC_ACT(BLK_TC_FS);
- __blk_add_trace(bt, rq->hard_sector, rq->hard_nr_sectors << 9,
rw, what, rq->errors, 0, NULL);
+ __blk_add_trace(bt, rq->hard_sector, rq->hard_nr_sectors << 9,
rw, what, rq->errors, ioprio, 0, NULL);
}
}
@@ -224,11 +226,12 @@ static inline void blk_add_trace_bio(str
u32 what)
{
struct blk_trace *bt = q->blk_trace;
+ unsigned short ioprio = bio_get_ioprio(bio);
if (likely(!bt))
return;
- __blk_add_trace(bt, bio->bi_sector, bio->bi_size, bio->bi_rw, what,
!bio_flagged(bio, BIO_UPTODATE), 0, NULL);
+ __blk_add_trace(bt, bio->bi_sector, bio->bi_size, bio->bi_rw, what,
!bio_flagged(bio, BIO_UPTODATE), ioprio, 0, NULL);
}
/**
@@ -253,7 +256,7 @@ static inline void blk_add_trace_generic
if (bio)
blk_add_trace_bio(q, bio, what);
else
- __blk_add_trace(bt, 0, 0, rw, what, 0, 0, NULL);
+ __blk_add_trace(bt, 0, 0, rw, what, 0, 0, 0, NULL);
}
/**
@@ -274,14 +277,17 @@ static inline void blk_add_trace_pdu_int
{
struct blk_trace *bt = q->blk_trace;
__be64 rpdu = cpu_to_be64(pdu);
+ unsigned short ioprio = 0;
if (likely(!bt))
return;
- if (bio)
- __blk_add_trace(bt, bio->bi_sector, bio->bi_size, bio->bi_rw, what,
!bio_flagged(bio, BIO_UPTODATE), sizeof(rpdu), &rpdu);
- else
- __blk_add_trace(bt, 0, 0, 0, what, 0, sizeof(rpdu), &rpdu);
+ if (bio) {
+ ioprio = bio_get_ioprio(bio);
+ __blk_add_trace(bt, bio->bi_sector, bio->bi_size, bio->bi_rw, what,
!bio_flagged(bio, BIO_UPTODATE), ioprio, sizeof(rpdu), &rpdu);
+ } else {
+ __blk_add_trace(bt, 0, 0, 0, what, 0, ioprio, sizeof(rpdu), &rpdu);
+ }
}
/**
@@ -302,6 +308,7 @@ static inline void blk_add_trace_remap(s
{
struct blk_trace *bt = q->blk_trace;
struct blk_io_trace_remap r;
+ unsigned short ioprio = bio_get_ioprio(bio);
if (likely(!bt))
return;
@@ -310,7 +317,7 @@ static inline void blk_add_trace_remap(s
r.device_from = cpu_to_be32(bio->bi_bdev->bd_dev);
r.sector = cpu_to_be64(to);
- __blk_add_trace(bt, from, bio->bi_size, bio->bi_rw, BLK_TA_REMAP,
!bio_flagged(bio, BIO_UPTODATE), sizeof(r), &r);
+ __blk_add_trace(bt, from, bio->bi_size, bio->bi_rw, BLK_TA_REMAP,
!bio_flagged(bio, BIO_UPTODATE), ioprio, sizeof(r), &r);
}
extern int blk_trace_setup(struct request_queue *q, char *name, dev_t dev,