direct-io: btrfs: avoid splitting dio requests for non-btrfs filesystems
Subsystems (3)
RFC PATCH 1/1 direct-io: btrfs: avoid splitting dio requests for non-btrfs filesystems
Hi,
this is about an issue newer kernels show, bysplitting direct I/O reque= sts
into 4k pieces to directly merge them in the Block Device Layer afterwa= rds.
If anyone is interested in own tests just use a simple command like dd if=3D/mnt/test/test-dd1 of=3D/dev/null iflag=3Ddirect bs=3D64k count= =3D1
in combination with blktrace.
The following patch is more a proposal for discussion than a solution, = well
thats what RFC's are about right.
I'm unsure about names, but also if the approach in general is the righ= t way.
It should apply to every 2.6.36 and 2.6.37 kernel.
I put everyone on CC who was involved in the patches leading to the cur= rent
behavior.
Gr=FCsse / regards,
Christian Ehrhardt
IBM Linux Technology Center, System z Linux Performance=20
--- cut here ---
Subject: [RFC][PATCH] direct-io: btrfs: avoid splitting dio requests fo= r non-btrfs filesystems
=46rom: Christian Ehrhardt <[email protected]>
Commit c2c6ca41 by Josef Bacik <[email protected]> caused all direct I/O= 's to
be split into 4k requests before arriving in the block device layer. This was later on partially fixed by Jeff Moyer <[email protected]> in 7a801ac6.
Jeffs fix improved the situation a lot, but eventually it still splits = I/Os
for non-btrfs file systems as well were it wouldn't have to.
Eventually in my example on a ext2 filesystem it splits it every 4Mb wh= eredio->boundary is evaluated to true.
In blktrace this looks like:
dd-910 [002] 38.762523: 94,8 A R 131264 + 8 <-=
(94,9) 131072
dd-910 [002] 38.762531: 94,8 Q R 131264 + 8 [d=
d]
dd-910 [002] 38.762535: 94,8 G R 131264 + 8 [d=
d]
dd-910 [002] 38.762537: 94,8 P N [dd]
dd-910 [002] 38.762539: 94,8 I R 131264 + 8 [d=
d]
dd-910 [002] 38.762544: 94,8 A R 131272 + 8 <-=
(94,9) 131080
dd-910 [002] 38.762544: 94,8 Q R 131272 + 8 [d=
d]
dd-910 [002] 38.762546: 94,8 M R 131272 + 8 [d=
d]
dd-910 [002] 38.762550: 94,8 A R 131280 + 8 <-=
(94,9) 131088
dd-910 [002] 38.762551: 94,8 Q R 131280 + 8 [d=
d]
dd-910 [002] 38.762551: 94,8 M R 131280 + 8 [d=
d]
dd-910 [002] 38.762556: 94,8 A R 131288 + 8 <-=
(94,9) 131096
dd-910 [002] 38.762557: 94,8 Q R 131288 + 8 [d=
d]
dd-910 [002] 38.762557: 94,8 M R 131288 + 8 [d=
d]
dd-910 [002] 38.762562: 94,8 A R 131296 + 8 <-=
(94,9) 131104
dd-910 [002] 38.762563: 94,8 Q R 131296 + 8 [d=
d]
dd-910 [002] 38.762564: 94,8 M R 131296 + 8 [d=
d]
dd-910 [002] 38.762568: 94,8 A R 131304 + 8 <-=
(94,9) 131112
dd-910 [002] 38.762569: 94,8 Q R 131304 + 8 [d=
d]
dd-910 [002] 38.762570: 94,8 M R 131304 + 8 [d=
d]
dd-910 [002] 38.762577: 94,8 A R 131312 + 8 <-=
(94,9) 131120
dd-910 [002] 38.762578: 94,8 Q R 131312 + 8 [d=
d]
dd-910 [002] 38.762579: 94,8 M R 131312 + 8 [d=
d]
dd-910 [002] 38.762584: 94,8 A R 131320 + 8 <-=
(94,9) 131128
dd-910 [002] 38.762584: 94,8 Q R 131320 + 8 [d=
d]
dd-910 [002] 38.762585: 94,8 M R 131320 + 8 [d=
d]
dd-910 [002] 38.762590: 94,8 A R 131328 + 8 <-=
(94,9) 131136
dd-910 [002] 38.762590: 94,8 Q R 131328 + 8 [d=
d]
dd-910 [002] 38.762591: 94,8 M R 131328 + 8 [d=
d]
dd-910 [002] 38.762596: 94,8 A R 131336 + 8 <-=
(94,9) 131144
dd-910 [002] 38.762597: 94,8 Q R 131336 + 8 [d=
d]
dd-910 [002] 38.762598: 94,8 M R 131336 + 8 [d=
d]
dd-910 [002] 38.762605: 94,8 A R 131344 + 16 <=
- (94,9) 131152
dd-910 [002] 38.762607: 94,8 Q R 131344 + 16 [=
dd]
dd-910 [002] 38.762608: 94,8 M R 131344 + 16 [=
dd]
dd-910 [002] 38.762611: 94,8 A R 131368 + 32 <=
- (94,9) 131176
dd-910 [002] 38.762612: 94,8 Q R 131368 + 32 [=
dd]
dd-910 [002] 38.762616: 94,8 G R 131368 + 32 [=
dd]
dd-910 [002] 38.762617: 94,8 I R 131368 + 32 [=
dd]
dd-910 [002] 38.762619: 94,8 U N [dd] 2
dd-910 [002] 38.762621: 94,8 D R 131264 + 96 [=
dd]
dd-910 [002] 38.762625: 94,8 D R 131368 + 32 [=
dd]
<idle>-0 [012] 38.763363: 94,8 C R 131264 + 96 [=
0]=20
<idle>-0 [015] 38.763797: 94,8 C R 131368 + 32 [=0]
The usual behavior before both commits was:
dd-919 [002] 37.513685: 94,8 A R 7824 + 96 <- =
(94,9) 7632
dd-919 [002] 37.513693: 94,8 Q R 7824 + 96 [dd=
]
dd-919 [002] 37.513697: 94,8 G R 7824 + 96 [dd=
]
dd-919 [002] 37.513700: 94,8 P N [dd]
dd-919 [002] 37.513701: 94,8 I R 7824 + 96 [dd=
]
dd-919 [002] 37.513794: 94,8 A R 7928 + 32 <- =
(94,9) 7736
dd-919 [002] 37.513795: 94,8 Q R 7928 + 32 [dd=
]
dd-919 [002] 37.513800: 94,8 G R 7928 + 32 [dd=
]
dd-919 [002] 37.513802: 94,8 I R 7928 + 32 [dd=
]
dd-919 [002] 37.513804: 94,8 U N [dd] 2
dd-919 [002] 37.513806: 94,8 D R 7824 + 96 [dd=
]
dd-919 [002] 37.513810: 94,8 D R 7928 + 32 [dd=
]
<idle>-0 [011] 37.514362: 94,8 C R 7824 + 96 [0]=
=20
<idle>-0 [014] 37.514728: 94,8 C R 7928 + 32 [0]That remaining split is cause by the test for:
"dio->final_block_in_bio !=3D dio->cur_page_block".
As this was in the code for a long time I just assume it is right.
So eventually for the 64k request in the example this patch improves th= e
effective submissions that get to the block device layer from:
10x4k, 1x8k, 1x16k to 1x48k & 1x16k which is much better.
Througput impact is small, but in terms of cpu consumption this is visi= ble
by a single digit percentage depending on the incoming request size.
The solution looking for comments or alternatives in this RFC patch add= s a new
kiocb flag that let filesystems specify if they need these workaround t= o
separate meta data reads - if not, like all pre-btrfs filesystems the d= io code
doesn't have to cause this extra work.
---
[diffstat]
fs/btrfs/inode.c | 4 ++++ fs/direct-io.c | 9 ++++++++- include/linux/aio.h | 5 +++++ 3 files changed, 17 insertions(+), 1 deletion(-)
[diff]
fs/btrfs/inode.c +1 -0
@@ -37,6 +37,7 @@#include <linux/posix_acl.h> #include <linux/falloc.h> #include <linux/slab.h> +#include <linux/aio.h> #include "compat.h" #include "ctree.h" #include "disk-io.h"
@@ -5822,6 +5823,9 @@ static ssize_t btrfs_direct_IO(int rw, struct kio=
cb *iocb,
free_extent_state(cached_state);
cached_state =3D NULL;
=20
+ /* btrfs cannot handle logically non-contiguous requests */
+ kiocb_set_separate_meta_reads(iocb);
+
ret =3D __blockdev_direct_IO(rw, iocb, inode,
BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev, iov, offset, nr_segs, btrfs_get_blocks_direct, NULL,
fs/direct-io.c +1 -0
@@ -35,6 +35,7 @@#include <linux/buffer_head.h> #include <linux/rwsem.h> #include <linux/uio.h> +#include <linux/aio.h> #include <asm/atomic.h>
=20
/*
@@ -79,6 +80,7 @@ struct dio {
sector_t final_block_in_request;/* doesn't change */
unsigned first_block_in_page; /* doesn't change, Used only once */
int boundary; /* prev block is at a boundary */
+ int separate_meta_reads; /* separate in I/O submission */
int reap_counter; /* rate limit reaping */
get_block_t *get_block; /* block mapping function */
dio_iodone_t *end_io; /* IO completion function */
@@ -659,7 +661,7 @@ static int dio_send_cur_page(struct dio *dio)
* Submit now if the underlying fs is about to perform a
* metadata read
*/
- else if (dio->boundary)
+ else if (dio->separate_meta_reads && dio->boundary)
dio_bio_submit(dio);
}
=20
@@ -1245,6 +1247,11 @@ __blockdev_direct_IO(int rw, struct kiocb *iocb,=
struct inode *inode,
dio->is_async =3D !is_sync_kiocb(iocb) && !((rw & WRITE) &&
(end > i_size_read(inode)));
=20
+ /*
+ * some filesystems e.g. btrfs need to separate metadata read
+ */
+ dio->separate_meta_reads =3D kiocb_needs_separate_meta_reads(iocb);
+
retval =3D direct_io_worker(rw, iocb, inode, iov, offset,
nr_segs, blkbits, get_block, end_io, submit_io, dio);
include/linux/aio.h +2 -0
@@ -34,6 +34,8 @@ struct kioctx;/* #define KIF_LOCKED 0 */ #define KIF_KICKED 1#define KIF_CANCELLED 2 +/* to separate meta reads */ +#define KIF_SEPARATE_META 3
=20
#define kiocbTryLock(iocb) test_and_set_bit(KIF_LOCKED, &(iocb)->ki_fl=
ags)
#define kiocbTryKick(iocb) test_and_set_bit(KIF_KICKED, &(iocb)->ki_fl=
ags)
@@ -50,6 +52,9 @@ struct kioctx;
#define kiocbIsKicked(iocb) test_bit(KIF_KICKED, &(iocb)->ki_flags)
#define kiocbIsCancelled(iocb) test_bit(KIF_CANCELLED, &(iocb)->ki_fla=
gs)
=20
+#define kiocb_set_separate_meta_reads(iocb) set_bit(KIF_SEPARATE_META,=
&(iocb)->ki_flags)
+#define kiocb_needs_separate_meta_reads(iocb) (KIF_SEPARATE_META & (io=
cb)->ki_flags)
+
/* is there a better place to document function pointer methods? */ /**
* ki_retry - iocb forward progress callback
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" = in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Quoted 2 lines · Christian Ehrhardt
Hi,
this is about an issue newer kernels show, bysplitting direct I/O req=
uests
Quoted 1 line
into 4k pieces to directly merge them in the Block Device Layer after=
wards.
Quoted 3 lines
=20
If anyone is interested in own tests just use a simple command like dd if=3D/mnt/test/test-dd1 of=3D/dev/null iflag=3Ddirect bs=3D64k cou=
nt=3D1
Quoted 3 lines
in combination with blktrace.
=20
The following patch is more a proposal for discussion than a solution=
, well
Quoted 2 lines
thats what RFC's are about right.
I'm unsure about names, but also if the approach in general is the ri=
ght way.
Quoted 3 lines
It should apply to every 2.6.36 and 2.6.37 kernel.
=20
I put everyone on CC who was involved in the patches leading to the c=
urrent
Quoted 9 lines
behavior.
=20
Gr=FCsse / regards,
Christian Ehrhardt
IBM Linux Technology Center, System z Linux Performance=20 =20
--- cut here ---
=20
Subject: [RFC][PATCH] direct-io: btrfs: avoid splitting dio requests =
for non-btrfs filesystems
Quoted 4 lines
=20
From: Christian Ehrhardt <[email protected]>
=20
Commitc2c6ca41by Josef Bacik <[email protected]> caused all direct I=
/O's to
Quoted 2 lines
be split into 4k requests before arriving in the block device layer. This was later on partially fixed by Jeff Moyer <[email protected]> i=
n
Quoted 3 lines
7a801ac6.
=20
Jeffs fix improved the situation a lot, but eventually it still split=
s I/Os
Quoted 3 lines
for non-btrfs file systems as well were it wouldn't have to. =20
Eventually in my example on a ext2 filesystem it splits it every 4Mb =
where
Quoted 4 lines
dio->boundaryis evaluated to true.
=20
In blktrace this looks like:dd-910 [002] 38.762523: 94,8 A R 131264 + 8 =
<- (94,9) 131072
Quoted 1 line
dd-910 [002] 38.762531: 94,8 Q R 131264 + 8 =
[dd]
Quoted 1 line
dd-910 [002] 38.762535: 94,8 G R 131264 + 8 =
[dd]
Quoted 2 lines
dd-910 [002] 38.762537: 94,8 P N [dd]
dd-910 [002] 38.762539: 94,8 I R 131264 + 8 =[dd]
Quoted 1 line
dd-910 [002] 38.762544: 94,8 A R 131272 + 8 =
<- (94,9) 131080
Quoted 1 line
dd-910 [002] 38.762544: 94,8 Q R 131272 + 8 =
[dd]
Quoted 1 line
dd-910 [002] 38.762546: 94,8 M R 131272 + 8 =
[dd]
Quoted 1 line
dd-910 [002] 38.762550: 94,8 A R 131280 + 8 =
<- (94,9) 131088
Quoted 1 line
dd-910 [002] 38.762551: 94,8 Q R 131280 + 8 =
[dd]
Quoted 1 line
dd-910 [002] 38.762551: 94,8 M R 131280 + 8 =
[dd]
Quoted 1 line
dd-910 [002] 38.762556: 94,8 A R 131288 + 8 =
<- (94,9) 131096
Quoted 1 line
dd-910 [002] 38.762557: 94,8 Q R 131288 + 8 =
[dd]
Quoted 1 line
dd-910 [002] 38.762557: 94,8 M R 131288 + 8 =
[dd]
Quoted 1 line
dd-910 [002] 38.762562: 94,8 A R 131296 + 8 =
<- (94,9) 131104
Quoted 1 line
dd-910 [002] 38.762563: 94,8 Q R 131296 + 8 =
[dd]
Quoted 1 line
dd-910 [002] 38.762564: 94,8 M R 131296 + 8 =
[dd]
Quoted 1 line
dd-910 [002] 38.762568: 94,8 A R 131304 + 8 =
<- (94,9) 131112
Quoted 1 line
dd-910 [002] 38.762569: 94,8 Q R 131304 + 8 =
[dd]
Quoted 1 line
dd-910 [002] 38.762570: 94,8 M R 131304 + 8 =
[dd]
Quoted 1 line
dd-910 [002] 38.762577: 94,8 A R 131312 + 8 =
<- (94,9) 131120
Quoted 1 line
dd-910 [002] 38.762578: 94,8 Q R 131312 + 8 =
[dd]
Quoted 1 line
dd-910 [002] 38.762579: 94,8 M R 131312 + 8 =
[dd]
Quoted 1 line
dd-910 [002] 38.762584: 94,8 A R 131320 + 8 =
<- (94,9) 131128
Quoted 1 line
dd-910 [002] 38.762584: 94,8 Q R 131320 + 8 =
[dd]
Quoted 1 line
dd-910 [002] 38.762585: 94,8 M R 131320 + 8 =
[dd]
Quoted 1 line
dd-910 [002] 38.762590: 94,8 A R 131328 + 8 =
<- (94,9) 131136
Quoted 1 line
dd-910 [002] 38.762590: 94,8 Q R 131328 + 8 =
[dd]
Quoted 1 line
dd-910 [002] 38.762591: 94,8 M R 131328 + 8 =
[dd]
Quoted 1 line
dd-910 [002] 38.762596: 94,8 A R 131336 + 8 =
<- (94,9) 131144
Quoted 1 line
dd-910 [002] 38.762597: 94,8 Q R 131336 + 8 =
[dd]
Quoted 1 line
dd-910 [002] 38.762598: 94,8 M R 131336 + 8 =
[dd]
Quoted 1 line
dd-910 [002] 38.762605: 94,8 A R 131344 + 16=
<- (94,9) 131152
Quoted 1 line
dd-910 [002] 38.762607: 94,8 Q R 131344 + 16=
[dd]
Quoted 1 line
dd-910 [002] 38.762608: 94,8 M R 131344 + 16=
[dd]
Quoted 1 line
dd-910 [002] 38.762611: 94,8 A R 131368 + 32=
<- (94,9) 131176
Quoted 1 line
dd-910 [002] 38.762612: 94,8 Q R 131368 + 32=
[dd]
Quoted 1 line
dd-910 [002] 38.762616: 94,8 G R 131368 + 32=
[dd]
Quoted 1 line
dd-910 [002] 38.762617: 94,8 I R 131368 + 32=
[dd]
Quoted 2 lines
dd-910 [002] 38.762619: 94,8 U N [dd] 2
dd-910 [002] 38.762621: 94,8 D R 131264 + 96=[dd]
Quoted 1 line
dd-910 [002] 38.762625: 94,8 D R 131368 + 32=
[dd]
Quoted 1 line
<idle>-0 [012] 38.763363: 94,8 C R 131264 + 96=
[0]=20
Quoted 1 line
<idle>-0 [015] 38.763797: 94,8 C R 131368 + 32=
[0]
Quoted 3 lines
=20
The usual behavior before both commits was:dd-919 [002] 37.513685: 94,8 A R 7824 + 96 <=
- (94,9) 7632
Quoted 1 line
dd-919 [002] 37.513693: 94,8 Q R 7824 + 96 [=
dd]
Quoted 1 line
dd-919 [002] 37.513697: 94,8 G R 7824 + 96 [=
dd]
Quoted 2 lines
dd-919 [002] 37.513700: 94,8 P N [dd]
dd-919 [002] 37.513701: 94,8 I R 7824 + 96 [=dd]
Quoted 1 line
dd-919 [002] 37.513794: 94,8 A R 7928 + 32 <=
- (94,9) 7736
Quoted 1 line
dd-919 [002] 37.513795: 94,8 Q R 7928 + 32 [=
dd]
Quoted 1 line
dd-919 [002] 37.513800: 94,8 G R 7928 + 32 [=
dd]
Quoted 1 line
dd-919 [002] 37.513802: 94,8 I R 7928 + 32 [=
dd]
Quoted 2 lines
dd-919 [002] 37.513804: 94,8 U N [dd] 2
dd-919 [002] 37.513806: 94,8 D R 7824 + 96 [=dd]
Quoted 1 line
dd-919 [002] 37.513810: 94,8 D R 7928 + 32 [=
dd]
Quoted 1 line
<idle>-0 [011] 37.514362: 94,8 C R 7824 + 96 [=
0]=20
Quoted 1 line
<idle>-0 [014] 37.514728: 94,8 C R 7928 + 32 [=
0]
Quoted 6 lines
=20
That remaining split is cause by the test for:
"dio->final_block_in_bio!=3Ddio->cur_page_block".
As this was in the code for a long time I just assume it is right. =20
So eventually for the 64k request in the example this patch improves =
the
Quoted 4 lines
effective submissions that get to the block device layer from:
10x4k, 1x8k, 1x16k to 1x48k & 1x16k which is much better. =20
Througput impact is small, but in terms of cpu consumption this is vi=
sible
Quoted 3 lines
by a single digit percentage depending on the incoming request size. =20
The solution looking for comments or alternatives in this RFC patch a=
dds a new
Quoted 1 line
kiocb flag that let filesystems specify if they need these workaround=
to
Quoted 1 line
separate meta data reads - if not, like all pre-btrfs filesystems the=
dio code
Quoted 2 lines
doesn't have to cause this extra work.
=20
So this brings up an important question, which is how badly do we want = to honorbuffer_boundary()? The whole reason I added the logical offset check w= as because
we ignore buffer_boundary() if there is no bio currently. So for BTRFS=
we would
do this
map extent
set buffer boundary
but if this is the first part of the IO and there isn't a bio already s= etup,dio_new_bio clears dio->boundary, so the next extent we would get may n= ot be
logically contiguous and hilarity would ensue. I chose not to fix the dio->boundary clearing because I was worried I would affect other fs's = workloads
(which I did anyway because of my bug). So maybe the right idea is to = rip out
my logical offset tests altogether and fix dio so we treat buffer_bound= ary()
like gospel. That way Btrfs can get what it needs without having this = weird
special code, and then we can look at how other fs's set buffer_boundar= y (I'm
pretty sure ext2/3 are the only ones) and make sure they are setting it=
when
they really mean to. Thanks,
Josef
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" = in
the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Quoted 6 lines · Josef Bacik
(which I did anyway because of my bug). So maybe the right idea is to rip out my logical offset tests altogether and fix dio so we treat
buffer_boundary()like gospel. That way Btrfs can get what it needs without having this weird special code, and then we can look at how other fs's setbuffer_boundary(I'm pretty sure ext2/3 are the only ones) and make sure they are setting it when they really mean to.
That sounds pretty reasonable to me. I really don't like the flag in the kiocb in this patch, and handling it as part of the get_blocks callback sounds much better to me. I don't know enough about the bounary blocks to know if we can reuse them - if we can it's perfect, if not another buffer flag seems like the way to go.