Re: [Qemu-devel] [PATCH 0/4] block: Vectored bdrv_writev_vmstate
Am 08.04.2013 um 18:18 hat Paolo Bonzini geschrieben: Il 08/04/2013 18:03, Stefan Hajnoczi ha scritto: Looks fine on the block side. Perhaps Orit, Juan, or Paolo can double-check the migration side. Reviewed-by: Stefan Hajnoczi stefa...@redhat.com Looks good, but given there are patches to fix it, I'm not sure it's worth the trouble... Shouldn't this perform even a little better? And we should get rid of non-vectored interfaces in the block layer anyway. Kevin
Re: [Qemu-devel] [PATCH 0/4] block: Vectored bdrv_writev_vmstate
Il 09/04/2013 10:04, Kevin Wolf ha scritto: Looks good, but given there are patches to fix it, I'm not sure it's worth the trouble... Shouldn't this perform even a little better? And we should get rid of non-vectored interfaces in the block layer anyway. Yes, if you have a very fast disk it should. Network throughput with Orit's patches jumped from 2.9 Gbps to 4.2 Gbps. But savevm is not live, so it is not as important to make it really fast. Anyhow, since the patches are there to fix both the cause and the symptom, there's really no reason to hold either series. Paolo
Re: [Qemu-devel] [PATCH 0/4] block: Vectored bdrv_writev_vmstate
On Tue, Apr 09, 2013 at 10:08:22AM +0200, Paolo Bonzini wrote: Il 09/04/2013 10:04, Kevin Wolf ha scritto: Looks good, but given there are patches to fix it, I'm not sure it's worth the trouble... Shouldn't this perform even a little better? And we should get rid of non-vectored interfaces in the block layer anyway. Yes, if you have a very fast disk it should. Network throughput with Orit's patches jumped from 2.9 Gbps to 4.2 Gbps. But savevm is not live, so it is not as important to make it really fast. Users have been asking about poor savevm/loadvm performance recently. I think anything we can do it improve it, especially since these patches already exist, is good. Stefan
Re: [Qemu-devel] [PATCH 0/4] block: Vectored bdrv_writev_vmstate
On Fri, Apr 05, 2013 at 09:27:52PM +0200, Kevin Wolf wrote: This gives us back reasonable savevm performance, which regressed in commit 500f0061. Kevin Wolf (4): block: Introduce bdrv_writev_vmstate savevm: Implement block_writev_buffer() block: Introduce bdrv_pwritev() for qcow2_save_vmstate qemu-iotests: A few more bdrv_pread/pwrite tests block.c | 105 +++--- block/qcow2.c | 6 +-- block/sheepdog.c | 13 -- include/block/block.h | 3 ++ include/block/block_int.h | 4 +- include/migration/qemu-file.h | 2 +- savevm.c | 25 -- tests/qemu-iotests/002| 13 ++ tests/qemu-iotests/002.out| 26 +++ 9 files changed, 157 insertions(+), 40 deletions(-) -- 1.8.1.4 Thanks, applied to my block tree: https://github.com/stefanha/qemu/commits/block Stefan
Re: [Qemu-devel] [PATCH 0/4] block: Vectored bdrv_writev_vmstate
On Fri, Apr 05, 2013 at 09:27:52PM +0200, Kevin Wolf wrote: This gives us back reasonable savevm performance, which regressed in commit 500f0061. Kevin Wolf (4): block: Introduce bdrv_writev_vmstate savevm: Implement block_writev_buffer() block: Introduce bdrv_pwritev() for qcow2_save_vmstate qemu-iotests: A few more bdrv_pread/pwrite tests block.c | 105 +++--- block/qcow2.c | 6 +-- block/sheepdog.c | 13 -- include/block/block.h | 3 ++ include/block/block_int.h | 4 +- include/migration/qemu-file.h | 2 +- savevm.c | 25 -- tests/qemu-iotests/002| 13 ++ tests/qemu-iotests/002.out| 26 +++ 9 files changed, 157 insertions(+), 40 deletions(-) -- 1.8.1.4 Looks fine on the block side. Perhaps Orit, Juan, or Paolo can double-check the migration side. Reviewed-by: Stefan Hajnoczi stefa...@redhat.com
Re: [Qemu-devel] [PATCH 0/4] block: Vectored bdrv_writev_vmstate
Il 08/04/2013 18:03, Stefan Hajnoczi ha scritto: Looks fine on the block side. Perhaps Orit, Juan, or Paolo can double-check the migration side. Reviewed-by: Stefan Hajnoczi stefa...@redhat.com Looks good, but given there are patches to fix it, I'm not sure it's worth the trouble... Paolo