From: Rusty Russell <rusty@rustcorp.com.au>
To: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>,
Dave Chinner <david@fromorbit.com>, Jens Axboe <axboe@kernel.dk>,
Minchan Kim <minchan@kernel.org>,
Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
Andrew Morton <akpm@linux-foundation.org>,
linux-mm <linux-mm@kvack.org>, "H. Peter Anvin" <hpa@zytor.com>,
Ingo Molnar <mingo@kernel.org>,
Peter Zijlstra <a.p.zijlstra@chello.nl>,
Mel Gorman <mgorman@suse.de>, Rik van Riel <riel@redhat.com>,
Johannes Weiner <hannes@cmpxchg.org>,
Hugh Dickins <hughd@google.com>,
Dave Hansen <dave.hansen@intel.com>,
Steven Rostedt <rostedt@goodmis.org>
Subject: Re: [PATCH 4/4] virtio_ring: unify direct/indirect code paths.
Date: Fri, 30 May 2014 12:07:44 +0930 [thread overview]
Message-ID: <87d2ew6lfr.fsf@rustcorp.com.au> (raw)
In-Reply-To: <20140529112905.GD30210@redhat.com>
"Michael S. Tsirkin" <mst@redhat.com> writes:
> On Thu, May 29, 2014 at 04:56:45PM +0930, Rusty Russell wrote:
>> virtqueue_add() populates the virtqueue descriptor table from the sgs
>> given. If it uses an indirect descriptor table, then it puts a single
>> descriptor in the descriptor table pointing to the kmalloc'ed indirect
>> table where the sg is populated.
>> + for (i = 0; i < total_sg; i++)
>> + desc[i].next = i+1;
>> + return desc;
>
> Hmm we are doing an extra walk over descriptors here.
> This might hurt performance esp for big descriptors.
Yes, this needs to be benchmarked; since it's cache hot my gut feel is
that it's a NOOP, but on modern machines my gut feel is always wrong.
>> + if (vq->indirect && total_sg > 1 && vq->vq.num_free)
>> + desc = alloc_indirect(total_sg, gfp);
>
> else desc = NULL will be a bit clearer won't it?
Agreed.
>> /* Update free pointer */
>> - vq->free_head = i;
>> + if (desc == vq->vring.desc)
>> + vq->free_head = i;
>> + else
>> + vq->free_head = vq->vring.desc[head].next;
>
> This one is slightly ugly isn't it?
Yes, but it avoided another variable, and I was originally aiming
at stack conservation. Turns out adding 'bool indirect' adds 32 bytes
more stack for gcc 4.6.4 :(
virtio_ring: minor neating
Before:
gcc 4.8.2: virtio_blk: stack used = 408
gcc 4.6.4: virtio_blk: stack used = 432
After:
gcc 4.8.2: virtio_blk: stack used = 408
gcc 4.6.4: virtio_blk: stack used = 464
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
index 3adf5978b92b..7a7849bc26af 100644
--- a/drivers/virtio/virtio_ring.c
+++ b/drivers/virtio/virtio_ring.c
@@ -141,9 +141,10 @@ static inline int virtqueue_add(struct virtqueue *_vq,
{
struct vring_virtqueue *vq = to_vvq(_vq);
struct scatterlist *sg;
- struct vring_desc *desc = NULL;
+ struct vring_desc *desc;
unsigned int i, n, avail, uninitialized_var(prev);
int head;
+ bool indirect;
START_USE(vq);
@@ -176,21 +177,25 @@ static inline int virtqueue_add(struct virtqueue *_vq,
* buffers, then go indirect. FIXME: tune this threshold */
if (vq->indirect && total_sg > 1 && vq->vq.num_free)
desc = alloc_indirect(total_sg, gfp);
+ else
+ desc = NULL;
if (desc) {
/* Use a single buffer which doesn't continue */
vq->vring.desc[head].flags = VRING_DESC_F_INDIRECT;
vq->vring.desc[head].addr = virt_to_phys(desc);
- /* avoid kmemleak false positive (tis hidden by virt_to_phys) */
+ /* avoid kmemleak false positive (hidden by virt_to_phys) */
kmemleak_ignore(desc);
vq->vring.desc[head].len = total_sg * sizeof(struct vring_desc);
/* Set up rest to use this indirect table. */
i = 0;
total_sg = 1;
+ indirect = true;
} else {
desc = vq->vring.desc;
i = head;
+ indirect = false;
}
if (vq->vq.num_free < total_sg) {
@@ -230,10 +235,10 @@ static inline int virtqueue_add(struct virtqueue *_vq,
desc[prev].flags &= ~VRING_DESC_F_NEXT;
/* Update free pointer */
- if (desc == vq->vring.desc)
- vq->free_head = i;
- else
+ if (indirect)
vq->free_head = vq->vring.desc[head].next;
+ else
+ vq->free_head = i;
/* Set token. */
vq->data[head] = data;
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
WARNING: multiple messages have this Message-ID (diff)
From: Rusty Russell <rusty@rustcorp.com.au>
To: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>,
Dave Chinner <david@fromorbit.com>, Jens Axboe <axboe@kernel.dk>,
Minchan Kim <minchan@kernel.org>,
Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
Andrew Morton <akpm@linux-foundation.org>,
linux-mm <linux-mm@kvack.org>, "H. Peter Anvin" <hpa@zytor.com>,
Ingo Molnar <mingo@kernel.org>,
Peter Zijlstra <a.p.zijlstra@chello.nl>,
Mel Gorman <mgorman@suse.de>, Rik van Riel <riel@redhat.com>,
Johannes Weiner <hannes@cmpxchg.org>,
Hugh Dickins <hughd@google.com>,
Dave Hansen <dave.hansen@intel.com>,
Steven Rostedt <rostedt@goodmis.org>
Subject: Re: [PATCH 4/4] virtio_ring: unify direct/indirect code paths.
Date: Fri, 30 May 2014 12:07:44 +0930 [thread overview]
Message-ID: <87d2ew6lfr.fsf@rustcorp.com.au> (raw)
In-Reply-To: <20140529112905.GD30210@redhat.com>
"Michael S. Tsirkin" <mst@redhat.com> writes:
> On Thu, May 29, 2014 at 04:56:45PM +0930, Rusty Russell wrote:
>> virtqueue_add() populates the virtqueue descriptor table from the sgs
>> given. If it uses an indirect descriptor table, then it puts a single
>> descriptor in the descriptor table pointing to the kmalloc'ed indirect
>> table where the sg is populated.
>> + for (i = 0; i < total_sg; i++)
>> + desc[i].next = i+1;
>> + return desc;
>
> Hmm we are doing an extra walk over descriptors here.
> This might hurt performance esp for big descriptors.
Yes, this needs to be benchmarked; since it's cache hot my gut feel is
that it's a NOOP, but on modern machines my gut feel is always wrong.
>> + if (vq->indirect && total_sg > 1 && vq->vq.num_free)
>> + desc = alloc_indirect(total_sg, gfp);
>
> else desc = NULL will be a bit clearer won't it?
Agreed.
>> /* Update free pointer */
>> - vq->free_head = i;
>> + if (desc == vq->vring.desc)
>> + vq->free_head = i;
>> + else
>> + vq->free_head = vq->vring.desc[head].next;
>
> This one is slightly ugly isn't it?
Yes, but it avoided another variable, and I was originally aiming
at stack conservation. Turns out adding 'bool indirect' adds 32 bytes
more stack for gcc 4.6.4 :(
virtio_ring: minor neating
Before:
gcc 4.8.2: virtio_blk: stack used = 408
gcc 4.6.4: virtio_blk: stack used = 432
After:
gcc 4.8.2: virtio_blk: stack used = 408
gcc 4.6.4: virtio_blk: stack used = 464
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
index 3adf5978b92b..7a7849bc26af 100644
--- a/drivers/virtio/virtio_ring.c
+++ b/drivers/virtio/virtio_ring.c
@@ -141,9 +141,10 @@ static inline int virtqueue_add(struct virtqueue *_vq,
{
struct vring_virtqueue *vq = to_vvq(_vq);
struct scatterlist *sg;
- struct vring_desc *desc = NULL;
+ struct vring_desc *desc;
unsigned int i, n, avail, uninitialized_var(prev);
int head;
+ bool indirect;
START_USE(vq);
@@ -176,21 +177,25 @@ static inline int virtqueue_add(struct virtqueue *_vq,
* buffers, then go indirect. FIXME: tune this threshold */
if (vq->indirect && total_sg > 1 && vq->vq.num_free)
desc = alloc_indirect(total_sg, gfp);
+ else
+ desc = NULL;
if (desc) {
/* Use a single buffer which doesn't continue */
vq->vring.desc[head].flags = VRING_DESC_F_INDIRECT;
vq->vring.desc[head].addr = virt_to_phys(desc);
- /* avoid kmemleak false positive (tis hidden by virt_to_phys) */
+ /* avoid kmemleak false positive (hidden by virt_to_phys) */
kmemleak_ignore(desc);
vq->vring.desc[head].len = total_sg * sizeof(struct vring_desc);
/* Set up rest to use this indirect table. */
i = 0;
total_sg = 1;
+ indirect = true;
} else {
desc = vq->vring.desc;
i = head;
+ indirect = false;
}
if (vq->vq.num_free < total_sg) {
@@ -230,10 +235,10 @@ static inline int virtqueue_add(struct virtqueue *_vq,
desc[prev].flags &= ~VRING_DESC_F_NEXT;
/* Update free pointer */
- if (desc == vq->vring.desc)
- vq->free_head = i;
- else
+ if (indirect)
vq->free_head = vq->vring.desc[head].next;
+ else
+ vq->free_head = i;
/* Set token. */
vq->data[head] = data;
next prev parent reply other threads:[~2014-05-30 6:02 UTC|newest]
Thread overview: 205+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-05-28 6:53 [PATCH 1/2] ftrace: print stack usage right before Oops Minchan Kim
2014-05-28 6:53 ` Minchan Kim
2014-05-28 6:53 ` [RFC 2/2] x86_64: expand kernel stack to 16K Minchan Kim
2014-05-28 6:53 ` Minchan Kim
2014-05-28 8:37 ` Dave Chinner
2014-05-28 8:37 ` Dave Chinner
2014-05-28 8:37 ` Dave Chinner
2014-05-28 9:13 ` Dave Chinner
2014-05-28 9:13 ` Dave Chinner
2014-05-28 9:13 ` Dave Chinner
2014-05-28 16:06 ` Johannes Weiner
2014-05-28 16:06 ` Johannes Weiner
2014-05-28 16:06 ` Johannes Weiner
2014-05-28 21:55 ` Dave Chinner
2014-05-28 21:55 ` Dave Chinner
2014-05-28 21:55 ` Dave Chinner
2014-05-29 6:06 ` Minchan Kim
2014-05-29 6:06 ` Minchan Kim
2014-05-29 6:06 ` Minchan Kim
2014-05-28 9:04 ` Michael S. Tsirkin
2014-05-28 9:04 ` Michael S. Tsirkin
2014-05-29 1:09 ` Minchan Kim
2014-05-29 2:44 ` Steven Rostedt
2014-05-29 2:44 ` Steven Rostedt
2014-05-29 4:11 ` Minchan Kim
2014-05-29 4:11 ` Minchan Kim
2014-05-29 2:47 ` Rusty Russell
2014-05-29 2:47 ` Rusty Russell
2014-05-29 4:10 ` virtio_ring stack usage Rusty Russell
2014-05-28 9:27 ` [RFC 2/2] x86_64: expand kernel stack to 16K Borislav Petkov
2014-05-29 13:23 ` One Thousand Gnomes
2014-05-29 13:23 ` One Thousand Gnomes
2014-05-28 14:14 ` Steven Rostedt
2014-05-28 14:14 ` Steven Rostedt
2014-05-28 14:23 ` H. Peter Anvin
2014-05-28 14:23 ` H. Peter Anvin
2014-05-28 22:11 ` Dave Chinner
2014-05-28 22:11 ` Dave Chinner
2014-05-28 22:42 ` H. Peter Anvin
2014-05-28 22:42 ` H. Peter Anvin
2014-05-28 23:17 ` Dave Chinner
2014-05-28 23:17 ` Dave Chinner
2014-05-28 23:21 ` H. Peter Anvin
2014-05-28 23:21 ` H. Peter Anvin
2014-05-28 15:43 ` Richard Weinberger
2014-05-28 15:43 ` Richard Weinberger
2014-05-28 16:08 ` Steven Rostedt
2014-05-28 16:08 ` Steven Rostedt
2014-05-28 16:11 ` Richard Weinberger
2014-05-28 16:11 ` Richard Weinberger
2014-05-28 16:13 ` Linus Torvalds
2014-05-28 16:13 ` Linus Torvalds
2014-05-28 16:09 ` Linus Torvalds
2014-05-28 16:09 ` Linus Torvalds
2014-05-28 22:31 ` Dave Chinner
2014-05-28 22:31 ` Dave Chinner
2014-05-28 22:41 ` Linus Torvalds
2014-05-28 22:41 ` Linus Torvalds
2014-05-29 1:30 ` Dave Chinner
2014-05-29 1:30 ` Dave Chinner
2014-05-29 1:58 ` Dave Chinner
2014-05-29 1:58 ` Dave Chinner
2014-05-29 2:51 ` Linus Torvalds
2014-05-29 2:51 ` Linus Torvalds
2014-05-29 23:36 ` Minchan Kim
2014-05-29 23:36 ` Minchan Kim
2014-05-30 0:05 ` Linus Torvalds
2014-05-30 0:20 ` Minchan Kim
2014-05-30 0:20 ` Minchan Kim
2014-05-30 0:31 ` Linus Torvalds
2014-05-30 0:31 ` Linus Torvalds
2014-05-30 0:50 ` Minchan Kim
2014-05-30 0:50 ` Minchan Kim
2014-05-30 1:24 ` Linus Torvalds
2014-05-30 1:24 ` Linus Torvalds
2014-05-30 1:58 ` Dave Chinner
2014-05-30 1:58 ` Dave Chinner
2014-05-30 2:13 ` Linus Torvalds
2014-05-30 2:13 ` Linus Torvalds
2014-05-30 6:21 ` Minchan Kim
2014-05-30 6:21 ` Minchan Kim
2014-05-30 1:30 ` Linus Torvalds
2014-05-30 1:30 ` Linus Torvalds
2014-05-30 0:15 ` Dave Chinner
2014-05-30 0:15 ` Dave Chinner
2014-05-30 2:12 ` Minchan Kim
2014-05-30 2:12 ` Minchan Kim
2014-05-30 4:37 ` Linus Torvalds
2014-05-30 4:37 ` Linus Torvalds
2014-05-31 1:45 ` Linus Torvalds
2014-05-31 1:45 ` Linus Torvalds
2014-05-30 6:12 ` Minchan Kim
2014-05-30 6:12 ` Minchan Kim
2014-06-03 13:28 ` Rasmus Villemoes
2014-06-03 13:28 ` Rasmus Villemoes
2014-06-03 19:04 ` Linus Torvalds
2014-06-03 19:04 ` Linus Torvalds
2014-06-10 12:29 ` [PATCH 0/2] Per-task wait_queue_t Rasmus Villemoes
2014-06-10 12:29 ` [PATCH 1/2] wait: Introduce per-task wait_queue_t Rasmus Villemoes
2014-06-11 15:16 ` Oleg Nesterov
2014-06-10 12:29 ` [PATCH 2/2] wait: Use the per-task wait_queue_t in ___wait_event macro Rasmus Villemoes
2014-06-10 15:50 ` [PATCH 0/2] Per-task wait_queue_t Peter Zijlstra
2014-06-12 21:46 ` Rasmus Villemoes
2014-05-29 2:42 ` [RFC 2/2] x86_64: expand kernel stack to 16K Linus Torvalds
2014-05-29 2:42 ` Linus Torvalds
2014-05-29 5:14 ` H. Peter Anvin
2014-05-29 5:14 ` H. Peter Anvin
2014-05-29 6:01 ` Rusty Russell
2014-05-29 6:01 ` Rusty Russell
2014-05-29 7:26 ` virtio ring cleanups, which save stack on older gcc Rusty Russell
2014-05-29 7:26 ` Rusty Russell
2014-05-29 7:26 ` [PATCH 1/4] Hack: measure stack taken by vring from virtio_blk Rusty Russell
2014-05-29 7:26 ` Rusty Russell
2014-05-29 15:39 ` Linus Torvalds
2014-05-29 15:39 ` Linus Torvalds
2014-05-29 7:26 ` [PATCH 2/4] virtio_net: pass well-formed sg to virtqueue_add_inbuf() Rusty Russell
2014-05-29 7:26 ` Rusty Russell
2014-05-29 10:07 ` Michael S. Tsirkin
2014-05-29 10:07 ` Michael S. Tsirkin
2014-05-29 7:26 ` [PATCH 3/4] virtio_ring: assume sgs are always well-formed Rusty Russell
2014-05-29 7:26 ` Rusty Russell
2014-05-29 11:18 ` Michael S. Tsirkin
2014-05-29 11:18 ` Michael S. Tsirkin
2014-05-29 7:26 ` [PATCH 4/4] virtio_ring: unify direct/indirect code paths Rusty Russell
2014-05-29 7:26 ` Rusty Russell
2014-05-29 7:52 ` Peter Zijlstra
2014-05-29 11:05 ` Rusty Russell
2014-05-29 11:05 ` Rusty Russell
2014-05-29 11:33 ` Michael S. Tsirkin
2014-05-29 11:33 ` Michael S. Tsirkin
2014-05-29 11:29 ` Michael S. Tsirkin
2014-05-29 11:29 ` Michael S. Tsirkin
2014-05-30 2:37 ` Rusty Russell [this message]
2014-05-30 2:37 ` Rusty Russell
2014-05-30 6:21 ` Rusty Russell
2014-05-29 7:41 ` virtio ring cleanups, which save stack on older gcc Minchan Kim
2014-05-29 7:41 ` Minchan Kim
2014-05-29 10:39 ` Dave Chinner
2014-05-29 10:39 ` Dave Chinner
2014-05-29 11:08 ` Rusty Russell
2014-05-29 11:08 ` Rusty Russell
2014-05-29 23:45 ` Minchan Kim
2014-05-29 23:45 ` Minchan Kim
2014-05-30 1:06 ` Minchan Kim
2014-05-30 1:06 ` Minchan Kim
2014-05-30 6:56 ` Rusty Russell
2014-05-30 6:56 ` Rusty Russell
2014-05-29 7:26 ` [RFC 2/2] x86_64: expand kernel stack to 16K Dave Chinner
2014-05-29 7:26 ` Dave Chinner
2014-05-29 15:24 ` Linus Torvalds
2014-05-29 15:24 ` Linus Torvalds
2014-05-29 23:40 ` Minchan Kim
2014-05-29 23:40 ` Minchan Kim
2014-05-29 23:53 ` Dave Chinner
2014-05-29 23:53 ` Dave Chinner
2014-05-30 0:06 ` Dave Jones
2014-05-30 0:06 ` Dave Jones
2014-05-30 0:21 ` Dave Chinner
2014-05-30 0:21 ` Dave Chinner
2014-05-30 0:29 ` Dave Jones
2014-05-30 0:29 ` Dave Jones
2014-05-30 0:32 ` Minchan Kim
2014-05-30 0:32 ` Minchan Kim
2014-05-30 1:34 ` Dave Chinner
2014-05-30 1:34 ` Dave Chinner
2014-05-30 15:25 ` H. Peter Anvin
2014-05-30 15:25 ` H. Peter Anvin
2014-05-30 15:41 ` Linus Torvalds
2014-05-30 15:41 ` Linus Torvalds
2014-05-30 15:52 ` H. Peter Anvin
2014-05-30 15:52 ` H. Peter Anvin
2014-05-30 16:06 ` Linus Torvalds
2014-05-30 16:06 ` Linus Torvalds
2014-05-30 17:24 ` Dave Hansen
2014-05-30 17:24 ` Dave Hansen
2014-05-30 18:12 ` H. Peter Anvin
2014-05-30 18:12 ` H. Peter Anvin
2014-10-21 2:00 ` Dave Jones
2014-10-21 4:59 ` Andy Lutomirski
2014-05-30 9:48 ` Richard Weinberger
2014-05-30 9:48 ` Richard Weinberger
2014-05-30 15:36 ` Linus Torvalds
2014-05-30 15:36 ` Linus Torvalds
2014-05-31 2:06 ` Jens Axboe
2014-05-31 2:06 ` Jens Axboe
2014-06-02 22:59 ` Dave Chinner
2014-06-02 22:59 ` Dave Chinner
2014-06-03 13:02 ` Konstantin Khlebnikov
2014-06-03 13:02 ` Konstantin Khlebnikov
2014-05-29 3:46 ` Minchan Kim
2014-05-29 3:46 ` Minchan Kim
2014-05-29 4:13 ` Linus Torvalds
2014-05-29 4:13 ` Linus Torvalds
2014-05-29 5:10 ` Minchan Kim
2014-05-29 5:10 ` Minchan Kim
2014-05-30 21:23 ` Andi Kleen
2014-05-30 21:23 ` Andi Kleen
2014-05-28 16:18 ` [PATCH 1/2] ftrace: print stack usage right before Oops Steven Rostedt
2014-05-28 16:18 ` Steven Rostedt
2014-05-29 3:52 ` Minchan Kim
2014-05-29 3:52 ` Minchan Kim
2014-05-29 3:01 ` Steven Rostedt
2014-05-29 3:01 ` Steven Rostedt
2014-05-29 3:49 ` Minchan Kim
2014-05-29 3:49 ` Minchan Kim
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=87d2ew6lfr.fsf@rustcorp.com.au \
--to=rusty@rustcorp.com.au \
--cc=a.p.zijlstra@chello.nl \
--cc=akpm@linux-foundation.org \
--cc=axboe@kernel.dk \
--cc=dave.hansen@intel.com \
--cc=david@fromorbit.com \
--cc=hannes@cmpxchg.org \
--cc=hpa@zytor.com \
--cc=hughd@google.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=mgorman@suse.de \
--cc=minchan@kernel.org \
--cc=mingo@kernel.org \
--cc=mst@redhat.com \
--cc=riel@redhat.com \
--cc=rostedt@goodmis.org \
--cc=torvalds@linux-foundation.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.