similar to: [PATCH net 0/3] vhost fixes for 3.14, -stable

vhost checked the counter within the refcnt before decrementing. It really wanted to know that it is the one that has the last reference, as a way to batch freeing resources a bit more efficiently. Note: we only let refcount go to 0 on device release. This works well but we now access the ref counter twice so there's a race: all users might see a high count and decide to defer freeing

vhost checked the counter within the refcnt before decrementing. It really wanted to know that it is the one that has the last reference, as a way to batch freeing resources a bit more efficiently. Note: we only let refcount go to 0 on device release. This works well but we now access the ref counter twice so there's a race: all users might see a high count and decide to defer freeing

vhost_zerocopy_callback accesses VQ right after it drops a ubuf reference. In theory, this could race with device removal which waits on the ubuf kref, and crash on use after free. Do all accesses within rcu read side critical section, and synchronize on release. Since callbacks are always invoked from bh, synchronize_rcu_bh seems enough and will help release complete a bit faster.

vhost_zerocopy_callback accesses VQ right after it drops a ubuf reference. In theory, this could race with device removal which waits on the ubuf kref, and crash on use after free. Do all accesses within rcu read side critical section, and synchronize on release. Since callbacks are always invoked from bh, synchronize_rcu_bh seems enough and will help release complete a bit faster.

Hi all: This series tries to unify and simplify vhost codes especially for zerocopy. With this series, 5% - 10% improvement for per cpu throughput were seen during netperf guest sending test. Plase review. Changes from V1: - Fix the zerocopy enabling check by changing the check of upend_idx != done_idx to (upend_idx + 1) % UIO_MAXIOV == done_idx. - Switch to use put_user() in

Hi all: This series tries to unify and simplify vhost codes especially for zerocopy. With this series, 5% - 10% improvement for per cpu throughput were seen during netperf guest sending test. Plase review. Changes from V1: - Fix the zerocopy enabling check by changing the check of upend_idx != done_idx to (upend_idx + 1) % UIO_MAXIOV == done_idx. - Switch to use put_user() in

On 2013/8/30 12:29, Jason Wang wrote: > We used to poll vhost queue before making DMA is done, this is racy if vhost > thread were waked up before marking DMA is done which can result the signal to > be missed. Fix this by always poll the vhost thread before DMA is done. > > Signed-off-by: Jason Wang <jasowang at redhat.com> > --- > drivers/vhost/net.c | 9 +++++----

On 2013/8/30 12:29, Jason Wang wrote: > We used to poll vhost queue before making DMA is done, this is racy if vhost > thread were waked up before marking DMA is done which can result the signal to > be missed. Fix this by always poll the vhost thread before DMA is done. > > Signed-off-by: Jason Wang <jasowang at redhat.com> > --- > drivers/vhost/net.c | 9 +++++----

This series tries to unify and simplify vhost codes especially for zerocopy. With this series, 5% - 10% improvement for per cpu throughput were seen during netperf guest sending test. Plase review. Changes from V2: - Typo fixes and code style fix - Add performance gain in the commit log of patch 2/6 - Retest the update the result in patch 6/6 Changes from V1: - Fix the zerocopy enabling check

This series tries to unify and simplify vhost codes especially for zerocopy. With this series, 5% - 10% improvement for per cpu throughput were seen during netperf guest sending test. Plase review. Changes from V2: - Typo fixes and code style fix - Add performance gain in the commit log of patch 2/6 - Retest the update the result in patch 6/6 Changes from V1: - Fix the zerocopy enabling check

Hi all: This series tries to unify and simplify vhost codes especially for zerocopy. Plase review. Thanks Jason Wang (6): vhost_net: make vhost_zerocopy_signal_used() returns void vhost_net: use vhost_add_used_and_signal_n() in vhost_zerocopy_signal_used() vhost: switch to use vhost_add_used_n() vhost_net: determine whether or not to use zerocopy at one time vhost_net: poll vhost

Hi all: This series tries to unify and simplify vhost codes especially for zerocopy. Plase review. Thanks Jason Wang (6): vhost_net: make vhost_zerocopy_signal_used() returns void vhost_net: use vhost_add_used_and_signal_n() in vhost_zerocopy_signal_used() vhost: switch to use vhost_add_used_n() vhost_net: determine whether or not to use zerocopy at one time vhost_net: poll vhost

On Fri, Aug 16, 2013 at 01:16:29PM +0800, Jason Wang wrote: > We used to poll vhost queue before making DMA is done, this is racy if vhost > thread were waked up before marking DMA is done which can result the signal to > be missed. Fix this by always poll the vhost thread before DMA is done. > > Signed-off-by: Jason Wang <jasowang at redhat.com> Indeed, but

On Fri, Aug 16, 2013 at 01:16:29PM +0800, Jason Wang wrote: > We used to poll vhost queue before making DMA is done, this is racy if vhost > thread were waked up before marking DMA is done which can result the signal to > be missed. Fix this by always poll the vhost thread before DMA is done. > > Signed-off-by: Jason Wang <jasowang at redhat.com> Indeed, but

On Fri, 2013-08-30 at 12:29 +0800, Jason Wang wrote: > We used to poll vhost queue before making DMA is done, this is racy if vhost > thread were waked up before marking DMA is done which can result the signal to > be missed. Fix this by always poll the vhost thread before DMA is done. Does this bug only exist in net-next or is it older? Should the fix go to net and stable branches?

On Fri, 2013-08-30 at 12:29 +0800, Jason Wang wrote: > We used to poll vhost queue before making DMA is done, this is racy if vhost > thread were waked up before marking DMA is done which can result the signal to > be missed. Fix this by always poll the vhost thread before DMA is done. Does this bug only exist in net-next or is it older? Should the fix go to net and stable branches?

MST, This is on top of [PATCH 0/2] vhost-net fix ubuf. Asias He (11): vhost: Remove vhost_enable_zcopy in vhost.h vhost: Move VHOST_NET_FEATURES to net.c vhost: Make vhost a separate module vhost: Remove comments for hdr in vhost.h vhost: Simplify dev->vqs[i] access vhost-net: Cleanup vhost_ubuf and vhost_zcopy vhost-scsi: Remove unnecessary forward struct vhost_scsi declaration

MST, This is on top of [PATCH 0/2] vhost-net fix ubuf. Asias He (11): vhost: Remove vhost_enable_zcopy in vhost.h vhost: Move VHOST_NET_FEATURES to net.c vhost: Make vhost a separate module vhost: Remove comments for hdr in vhost.h vhost: Simplify dev->vqs[i] access vhost-net: Cleanup vhost_ubuf and vhost_zcopy vhost-scsi: Remove unnecessary forward struct vhost_scsi declaration

On Mon, Sep 02, 2013 at 04:40:59PM +0800, Jason Wang wrote: > Currently, even if the packet length is smaller than VHOST_GOODCOPY_LEN, if > upend_idx != done_idx we still set zcopy_used to true and rollback this choice > later. This could be avoided by determining zerocopy once by checking all > conditions at one time before. > > Signed-off-by: Jason Wang <jasowang at

On Mon, Sep 02, 2013 at 04:40:59PM +0800, Jason Wang wrote: > Currently, even if the packet length is smaller than VHOST_GOODCOPY_LEN, if > upend_idx != done_idx we still set zcopy_used to true and rollback this choice > later. This could be avoided by determining zerocopy once by checking all > conditions at one time before. > > Signed-off-by: Jason Wang <jasowang at