Skip to contentSkip to conversation

lguest & virtio

GIT PULL New Rusty Russell 2 participants 5 messages Sep 23, 2009 – Sep 28, 2009

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

The following changes since commit ebc79c4f8da0f92efa968e0328f32334a2ce80cf:
Linus Torvalds (1):

        Merge git://git.kernel.org/.../jaswinder/linux-2.6

are available in the git repository at:

ssh://master.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus.git master

Christoph Hellwig (1):
virtio_blk: add support for cache flush

Fernando Luis Vazquez Cao (2):
virtio: get rid of redundant VIRTIO_ID_9P definition
virtio: add virtio IDs file

Rusty Russell (6):
virtio_pci: minor MSI-X cleanups
virtio: make add_buf return capacity remaining
lguest: move panic notifier registration to its expected place.
lguest: use set_pte/set_pmd uniformly for real page table entries
lguest: use PGDIR_SHIFT for PAE code to allow different PAGE_OFFSET
lguest: don't force VIRTIO_F_NOTIFY_ON_EMPTY

Xiao Guangrong (1):
lguest: cleanup for map_switcher()

 Documentation/lguest/lguest.c       |   26 ++++++-
 arch/x86/lguest/boot.c              |   10 +--
 drivers/block/virtio_blk.c          |   33 ++++++++--
 drivers/char/hw_random/virtio-rng.c |    3 +-
 drivers/char/virtio_console.c       |    5 +-
 drivers/lguest/core.c               |    5 +-
 drivers/lguest/page_tables.c        |   45 +++++-------
 drivers/net/virtio_net.c            |   15 ++--
 drivers/virtio/virtio_balloon.c     |    3 +-
 drivers/virtio/virtio_pci.c         |  125 +++++++++++++++++++---------------
 drivers/virtio/virtio_ring.c        |    6 ++-
 include/linux/virtio.h              |    2 +-
 include/linux/virtio_9p.h           |    2 -
 include/linux/virtio_balloon.h      |    3 -
 include/linux/virtio_blk.h          |   18 ++++-
 include/linux/virtio_console.h      |    3 -
 include/linux/virtio_ids.h          |   17 +++++
 include/linux/virtio_net.h          |    3 -
 include/linux/virtio_rng.h          |    3 -
 net/9p/trans_virtio.c               |    5 +-
 20 files changed, 199 insertions(+), 133 deletions(-)
 create mode 100644 include/linux/virtio_ids.h
#1 lore raw

What happend to the patch to drop the virt/non-rotational flag?

Quoted 40 lines · Rusty Russell

The following changes since commit ebc79c4f8da0f92efa968e0328f32334a2ce80cf:
Linus Torvalds (1):

        Merge git://git.kernel.org/.../jaswinder/linux-2.6

are available in the git repository at:

ssh://master.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus.git master

Christoph Hellwig (1):
virtio_blk: add support for cache flush

Fernando Luis Vazquez Cao (2):
virtio: get rid of redundant VIRTIO_ID_9P definition
virtio: add virtio IDs file

Rusty Russell (6):
virtio_pci: minor MSI-X cleanups
virtio: make add_buf return capacity remaining
lguest: move panic notifier registration to its expected place.
lguest: use set_pte/set_pmd uniformly for real page table entries
lguest: use PGDIR_SHIFT for PAE code to allow different PAGE_OFFSET
lguest: don't force VIRTIO_F_NOTIFY_ON_EMPTY

Xiao Guangrong (1):
lguest: cleanup for map_switcher()

 Documentation/lguest/lguest.c       |   26 ++++++-
 arch/x86/lguest/boot.c              |   10 +--
 drivers/block/virtio_blk.c          |   33 ++++++++--
 drivers/char/hw_random/virtio-rng.c |    3 +-
 drivers/char/virtio_console.c       |    5 +-
 drivers/lguest/core.c               |    5 +-
 drivers/lguest/page_tables.c        |   45 +++++-------
 drivers/net/virtio_net.c            |   15 ++--
 drivers/virtio/virtio_balloon.c     |    3 +-
 drivers/virtio/virtio_pci.c         |  125 +++++++++++++++++++---------------
 drivers/virtio/virtio_ring.c        |    6 ++-
 include/linux/virtio.h              |    2 +-
 include/linux/virtio_9p.h           |    2 -
 include/linux/virtio_balloon.h      |    3 -
Quoted 8 lines
 include/linux/virtio_blk.h          |   18 ++++-
 include/linux/virtio_console.h      |    3 -
 include/linux/virtio_ids.h          |   17 +++++
 include/linux/virtio_net.h          |    3 -
 include/linux/virtio_rng.h          |    3 -
 net/9p/trans_virtio.c               |    5 +-
 20 files changed, 199 insertions(+), 133 deletions(-)
 create mode 100644 include/linux/virtio_ids.h

---end quoted text---

#3 lore raw

PATCH 1/1 virtio_blk: revert QUEUE_FLAG_VIRT addition

In reply to Rusty Russell

From: Christoph Hellwig <[email protected]>

It seems like the addition of QUEUE_FLAG_VIRT caueses major performance regressions for Fedora users:

	https://bugzilla.redhat.com/show_bug.cgi?id=509383
	https://bugzilla.redhat.com/show_bug.cgi?id=505695

while I can't reproduce those extreme regressions myself I think the flag is wrong.

Rationale:

QUEUE_FLAG_VIRT expands to QUEUE_FLAG_NONROT which casus the queue unplugged immediately. This is not a good behaviour for at least qemu and kvm where we do have significant overhead for every I/O operations. Even with all the latested speeups (native AIO, MSI support, zero copy) we can only get native speed for up to 128kb I/O requests we already are down to 66% of native performance for 4kb requests even on my laptop running the Intel X25-M SSD for which the QUEUE_FLAG_NONROT was designed.
If we ever get virtio-blk overhead low enough that this flag makes sense it should only be set based on a feature flag set by the host.

---
 drivers/block/virtio_blk.c |    1 -
 1 file changed, 1 deletion(-)
drivers/block/virtio_blk.c +0 -1
Index: linux-2.6/drivers/block/virtio_blk.c
===================================================================
@@ -314,7 +314,6 @@ static int __devinit virtblk_probe(struc
 	}
 
 	vblk->disk->queue->queuedata = vblk;
-	queue_flag_set_unlocked(QUEUE_FLAG_VIRT, vblk->disk->queue);
 
 	if (index < 26) {
 		sprintf(vblk->disk->disk_name, "vd%c", 'a' + index % 26);
#2 lore raw
In reply to Christoph Hellwig
Quoted 1 line · Christoph Hellwig

What happend to the patch to drop the virt/non-rotational flag?

I pushed it back; I realize now you still wanted that one, it was "Subject: virtio_blk: set QUEUE_ORDERED_DRAIN by default" you wanted to reconsider.

Will send separately now...
Rusty.

#4 lore raw

PATCH 1/1 virtio_blk: revert QUEUE_FLAG_VIRT addition

From: Christoph Hellwig <[email protected]>

It seems like the addition of QUEUE_FLAG_VIRT caueses major performance regressions for Fedora users:

	https://bugzilla.redhat.com/show_bug.cgi?id=509383
	https://bugzilla.redhat.com/show_bug.cgi?id=505695

while I can't reproduce those extreme regressions myself I think the flag is wrong.

Rationale:

QUEUE_FLAG_VIRT expands to QUEUE_FLAG_NONROT which casus the queue unplugged immediately. This is not a good behaviour for at least qemu and kvm where we do have significant overhead for every I/O operations. Even with all the latested speeups (native AIO, MSI support, zero copy) we can only get native speed for up to 128kb I/O requests we already are down to 66% of native performance for 4kb requests even on my laptop running the Intel X25-M SSD for which the QUEUE_FLAG_NONROT was designed.
If we ever get virtio-blk overhead low enough that this flag makes sense it should only be set based on a feature flag set by the host.

---
 drivers/block/virtio_blk.c |    1 -
 1 file changed, 1 deletion(-)
drivers/block/virtio_blk.c +0 -1
Index: linux-2.6/drivers/block/virtio_blk.c
===================================================================
@@ -314,7 +314,6 @@ static int __devinit virtblk_probe(struc
 	}
 
 	vblk->disk->queue->queuedata = vblk;
-	queue_flag_set_unlocked(QUEUE_FLAG_VIRT, vblk->disk->queue);
 
 	if (index < 26) {
 		sprintf(vblk->disk->disk_name, "vd%c", 'a' + index % 26);