Support IOThread polling for PV shared rings
cover Support IOThread polling for PV shared rings
Currently xen-block uses an IOThread to handle AIO but the event channels are dealt with on QEMU's main thread. This series allows them to be dealt with in the same context.
Paul Durrant (3):
xen-bus: use a separate fd for each event channel
xen-bus: allow AioContext to be specified for each event channel xen-bus / xen-block: add support for event channel polling
hw/block/dataplane/xen-block.c | 19 +++---- hw/xen/xen-bus.c | 92 +++++++++++++++++++--------------- include/hw/xen/xen-bus.h | 9 ++-- 3 files changed, 66 insertions(+), 54 deletions(-)
---
Cc: Anthony Perard <[email protected]>
Cc: Kevin Wolf <[email protected]>
Cc: Max Reitz <[email protected]>
Cc: Stefan Hajnoczi <[email protected]>
Cc: Stefano Stabellini <[email protected]>
--
2.20.1.2.gb21ebb6
PATCH 1/3 xen-bus: use a separate fd for each event channel
To better support use of IOThread-s it will be necessary to be able to set the AioContext for each XenEventChannel and hence it is necessary to open a separate handle to libxenevtchan for each channel.
This patch stops using NotifierList for event channel callbacks, replacing that construct by a list of complete XenEventChannel structures. Each of these now has a xenevtchn_handle pointer in place of the single pointer previously held in the XenDevice structure. The individual handles are opened/closed in xen_device_bind/unbind_event_channel(), replacing the single open/close in xen_device_realize/unrealize().
NOTE: This patch does not add an AioContext parameter to
xen_device_bind_event_channel(). That will be done in a subsequent patch.
---
Cc: Stefano Stabellini <[email protected]>
Cc: Anthony Perard <[email protected]>
--- hw/xen/xen-bus.c | 79 ++++++++++++++++++++-------------------- include/hw/xen/xen-bus.h | 6 +-- 2 files changed, 42 insertions(+), 43 deletions(-)
hw/xen/xen-bus.c +40 -39
@@ -923,19 +923,22 @@ done: }struct XenEventChannel { +QLIST_ENTRY (XenEventChannel) list; +xenevtchn_handle *xeh;evtchn_port_t local_port;XenEventHandler handler;void *opaque; -Notifier notifier; }; -static void event_notify (Notifier *n,void *data) +static void xen_device_event (void *opaque) { -XenEventChannel *channel =container_of (n, XenEventChannel, notifier); -unsigned long port = (unsigned long )data; +XenEventChannel *channel = opaque; +unsigned long port =xenevtchn_pending (channel->xeh);if (port == channel->local_port) { channel->handler (channel->opaque); + +xenevtchn_unmask (channel->xeh, port); } }
@@ -947,24 +950,39 @@ XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev,XenEventChannel *channel =g_new0 (XenEventChannel,1 );xenevtchn_port_or_error_t local_port; - local_port =xenevtchn_bind_interdomain (xendev->xeh, + channel->xeh =xenevtchn_open (NULL ,0 ); +if (!channel->xeh) { +error_setg_errno (errp, errno,"failed xenevtchn_open" ); + goto fail; + } + + local_port =xenevtchn_bind_interdomain (channel->xeh, xendev->frontend_id, port);if (local_port <0 ) {error_setg_errno (errp, errno,"xenevtchn_bind_interdomain failed" ); - -g_free (channel); -return NULL ; + goto fail; } channel->local_port = local_port; channel->handler = handler; channel->opaque = opaque; - channel->notifier.notify = event_notify; -notifier_list_add (&xendev->event_notifiers, &channel->notifier); +qemu_set_fd_handler (xenevtchn_fd (channel->xeh), xen_device_event,NULL , + channel); + +QLIST_INSERT_HEAD (&xendev->event_channels, channel, list);return channel; + +fail: +if (channel->xeh) { +xenevtchn_close (channel->xeh); + } + +g_free (channel); + +return NULL ; } void xen_device_notify_event_channel(XenDevice *xendev,
@@ -976,7 +994,7 @@ void xen_device_notify_event_channel(XenDevice *xendev,return ; } -if (xenevtchn_notify (xendev->xeh, channel->local_port) <0 ) { +if (xenevtchn_notify (channel->xeh, channel->local_port) <0 ) {error_setg_errno (errp, errno,"xenevtchn_notify failed" ); } }
@@ -990,12 +1008,15 @@ void xen_device_unbind_event_channel(XenDevice *xendev,return ; } -notifier_remove (&channel->notifier); +QLIST_REMOVE (channel, list); -if (xenevtchn_unbind (xendev->xeh, channel->local_port) <0 ) { +qemu_set_fd_handler (xenevtchn_fd (channel->xeh),NULL ,NULL ,NULL ); + +if (xenevtchn_unbind (channel->xeh, channel->local_port) <0 ) {error_setg_errno (errp, errno,"xenevtchn_unbind failed" ); } +xenevtchn_close (channel->xeh);g_free (channel); }
@@ -1004,6 +1025,7 @@ static void xen_device_unrealize(DeviceState *dev, Error **errp)XenDevice *xendev =XEN_DEVICE (dev);XenDeviceClass *xendev_class =XEN_DEVICE_GET_CLASS (xendev);const char *type =object_get_typename (OBJECT (xendev)); +XenEventChannel *channel, *next;if (!xendev->name) {return ;
@@ -1020,15 +1042,14 @@ static void xen_device_unrealize(DeviceState *dev, Error **errp) xendev_class->unrealize (xendev, errp); } +/* Make sure all event channels are cleaned up */ + QLIST_FOREACH_SAFE(channel, &xendev->event_channels, list, next) { +xen_device_unbind_event_channel (xendev, channel,NULL ); + } +xen_device_frontend_destroy (xendev);xen_device_backend_destroy (xendev); -if (xendev->xeh) { -qemu_set_fd_handler (xenevtchn_fd (xendev->xeh),NULL ,NULL ,NULL ); -xenevtchn_close (xendev->xeh); - xendev->xeh =NULL ; - } -if (xendev->xgth) {xengnttab_close (xendev->xgth); xendev->xgth =NULL ;
@@ -1045,16 +1066,6 @@ static void xen_device_exit(Notifier *n, void *data)xen_device_unrealize (DEVICE (xendev), &error_abort); } -static void xen_device_event (void *opaque) -{ -XenDevice *xendev = opaque; -unsigned long port =xenevtchn_pending (xendev->xeh); - -notifier_list_notify (&xendev->event_notifiers, (void *)port); - -xenevtchn_unmask (xendev->xeh, port); -} -static void xen_device_realize (DeviceState *dev,Error **errp) {XenDevice *xendev =XEN_DEVICE (dev);
@@ -1095,16 +1106,6 @@ static void xen_device_realize(DeviceState *dev, Error **errp) xendev->feature_grant_copy = (xengnttab_grant_copy (xendev->xgth,0 ,NULL ) ==0 ); - xendev->xeh =xenevtchn_open (NULL ,0 ); -if (!xendev->xeh) { -error_setg_errno (errp, errno,"failed xenevtchn_open" ); - goto unrealize; - } - -notifier_list_init (&xendev->event_notifiers); -qemu_set_fd_handler (xenevtchn_fd (xendev->xeh), xen_device_event,NULL , - xendev); -xen_device_backend_create (xendev, &local_err);if (local_err) {error_propagate (errp, local_err);
include/hw/xen/xen-bus.h +2 -4
@@ -15,6 +15,7 @@typedef void (*XenWatchHandler )(void *opaque);typedef struct XenWatch XenWatch ; +typedef struct XenEventChannel XenEventChannel ; typedef struct XenDevice { DeviceState qdev;
@@ -28,8 +29,7 @@ typedef struct XenDevice {XenWatch *backend_online_watch;xengnttab_handle *xgth;bool feature_grant_copy; -xenevtchn_handle *xeh; -NotifierList event_notifiers; + QLIST_HEAD(, XenEventChannel) event_channels; } XenDevice;typedef char *(*XenDeviceGetName )(XenDevice *xendev,Error **errp);
@@ -119,8 +119,6 @@ void xen_device_copy_grant_refs(XenDevice *xendev, bool to_domain,XenDeviceGrantCopySegment segs[], unsigned int nr_segs, Error **errp); -typedef struct XenEventChannel XenEventChannel ; -typedef void (*XenEventHandler )(void *opaque); XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev,
Signature · 7 lines
2.20.1.2.gb21ebb6 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH 1/3] xen-bus: use a separate fd for each event channel
Quoted 11 lines · Paul Durrant
To better support use of IOThread-s it will be necessary to be able to set the AioContext for each XenEventChannel and hence it is necessary to open a separate handle to libxenevtchan for each channel.
This patch stops using NotifierList for event channel callbacks, replacing that construct by a list of complete XenEventChannel structures. Each of these now has a
xenevtchn_handlepointer in place of the single pointer previously held in the XenDevice structure. The individual handles are opened/closed in xen_device_bind/unbind_event_channel(), replacing the single open/close in xen_device_realize/unrealize().
Quoted 5 lines
NOTE: This patch does not add an AioContext parameter to
xen_device_bind_event_channel(). That will be done in a subsequent patch.Signed-off-by Paul Durrant <[email protected]>
There are a few places were I would have like to add an assert, but they can't be compiled-out in QEMU :-(.
Thanks,
Signature · 6 lines
Anthony PERARD _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [Qemu-devel] [PATCH 1/3] xen-bus: use a separate fd for each event channel
In reply to Paul DurrantQuoted 11 lines · Paul Durrant
To better support use of IOThread-s it will be necessary to be able to set the AioContext for each XenEventChannel and hence it is necessary to open a separate handle to libxenevtchan for each channel.
This patch stops using NotifierList for event channel callbacks, replacing that construct by a list of complete XenEventChannel structures. Each of these now has a
xenevtchn_handlepointer in place of the single pointer previously held in the XenDevice structure. The individual handles are opened/closed in xen_device_bind/unbind_event_channel(), replacing the single open/close in xen_device_realize/unrealize().
Quoted 5 lines
NOTE: This patch does not add an AioContext parameter to
xen_device_bind_event_channel(). That will be done in a subsequent patch.Signed-off-by Paul Durrant <[email protected]>
There are a few places were I would have like to add an assert, but they can't be compiled-out in QEMU :-(.
Thanks,
Signature · 1 line
Anthony PERARD
PATCH 2/3 xen-bus: allow AioContext to be specified for each event channel
In reply to Paul DurrantThis patch adds an AioContext parameter to xen_device_bind_event_channel() and then uses aio_set_fd_handler() to set the callback rather than qemu_set_fd_handler().
---
Cc: Stefano Stabellini <[email protected]>
Cc: Anthony Perard <[email protected]>
Cc: Stefan Hajnoczi <[email protected]>
Cc: Kevin Wolf <[email protected]>
Cc: Max Reitz <[email protected]>
--- hw/block/dataplane/xen-block.c | 2 +- hw/xen/xen-bus.c | 10 +++++++--- include/hw/xen/xen-bus.h | 1 + 3 files changed, 9 insertions(+), 4 deletions(-)
hw/block/dataplane/xen-block.c +1 -1
@@ -802,7 +802,7 @@ void xen_block_dataplane_start(XenBlockDataPlane *dataplane, } dataplane->event_channel = -xen_device_bind_event_channel (xendev, event_channel, +xen_device_bind_event_channel (xendev, dataplane->ctx, event_channel, xen_block_dataplane_event, dataplane, &local_err);if (local_err) {
hw/xen/xen-bus.c +7 -3
@@ -924,6 +924,7 @@ done:struct XenEventChannel {QLIST_ENTRY (XenEventChannel) list; +AioContext *ctx;xenevtchn_handle *xeh;evtchn_port_t local_port;XenEventHandler handler;
@@ -943,6 +944,7 @@ static void xen_device_event(void *opaque) }XenEventChannel *xen_device_bind_event_channel (XenDevice *xendev, +AioContext *ctx,unsigned int port,XenEventHandler handler,void *opaque,Error **errp)
@@ -968,8 +970,9 @@ XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, channel->handler = handler; channel->opaque = opaque; -qemu_set_fd_handler (xenevtchn_fd (channel->xeh), xen_device_event,NULL , - channel); + channel->ctx = ctx; +aio_set_fd_handler (channel->ctx,xenevtchn_fd (channel->xeh), false, + xen_device_event,NULL ,NULL , channel);QLIST_INSERT_HEAD (&xendev->event_channels, channel, list);
@@ -1010,7 +1013,8 @@ void xen_device_unbind_event_channel(XenDevice *xendev,QLIST_REMOVE (channel, list); -qemu_set_fd_handler (xenevtchn_fd (channel->xeh),NULL ,NULL ,NULL ); +aio_set_fd_handler (channel->ctx,xenevtchn_fd (channel->xeh), false, +NULL ,NULL ,NULL ,NULL );if (xenevtchn_unbind (channel->xeh, channel->local_port) <0 ) {error_setg_errno (errp, errno,"xenevtchn_unbind failed" );
include/hw/xen/xen-bus.h +1 -0
@@ -122,6 +122,7 @@ void xen_device_copy_grant_refs(XenDevice *xendev, bool to_domain,typedef void (*XenEventHandler )(void *opaque);XenEventChannel *xen_device_bind_event_channel (XenDevice *xendev, +AioContext *ctx,unsigned int port,XenEventHandler handler,void *opaque,Error **errp);
Signature · 7 lines
2.20.1.2.gb21ebb6 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
PATCH 2/3 xen-bus: allow AioContext to be specified for each event channel
In reply to Paul DurrantThis patch adds an AioContext parameter to xen_device_bind_event_channel() and then uses aio_set_fd_handler() to set the callback rather than qemu_set_fd_handler().
---
Cc: Stefano Stabellini <[email protected]>
Cc: Anthony Perard <[email protected]>
Cc: Stefan Hajnoczi <[email protected]>
Cc: Kevin Wolf <[email protected]>
Cc: Max Reitz <[email protected]>
--- hw/block/dataplane/xen-block.c | 2 +- hw/xen/xen-bus.c | 10 +++++++--- include/hw/xen/xen-bus.h | 1 + 3 files changed, 9 insertions(+), 4 deletions(-)
hw/block/dataplane/xen-block.c +1 -1
@@ -802,7 +802,7 @@ void xen_block_dataplane_start(XenBlockDataPlane *dataplane, } dataplane->event_channel = -xen_device_bind_event_channel (xendev, event_channel, +xen_device_bind_event_channel (xendev, dataplane->ctx, event_channel, xen_block_dataplane_event, dataplane, &local_err);if (local_err) {
hw/xen/xen-bus.c +7 -3
@@ -924,6 +924,7 @@ done:struct XenEventChannel {QLIST_ENTRY (XenEventChannel) list; +AioContext *ctx;xenevtchn_handle *xeh;evtchn_port_t local_port;XenEventHandler handler;
@@ -943,6 +944,7 @@ static void xen_device_event(void *opaque) }XenEventChannel *xen_device_bind_event_channel (XenDevice *xendev, +AioContext *ctx,unsigned int port,XenEventHandler handler,void *opaque,Error **errp)
@@ -968,8 +970,9 @@ XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, channel->handler = handler; channel->opaque = opaque; -qemu_set_fd_handler (xenevtchn_fd (channel->xeh), xen_device_event,NULL , - channel); + channel->ctx = ctx; +aio_set_fd_handler (channel->ctx,xenevtchn_fd (channel->xeh), false, + xen_device_event,NULL ,NULL , channel);QLIST_INSERT_HEAD (&xendev->event_channels, channel, list);
@@ -1010,7 +1013,8 @@ void xen_device_unbind_event_channel(XenDevice *xendev,QLIST_REMOVE (channel, list); -qemu_set_fd_handler (xenevtchn_fd (channel->xeh),NULL ,NULL ,NULL ); +aio_set_fd_handler (channel->ctx,xenevtchn_fd (channel->xeh), false, +NULL ,NULL ,NULL ,NULL );if (xenevtchn_unbind (channel->xeh, channel->local_port) <0 ) {error_setg_errno (errp, errno,"xenevtchn_unbind failed" );
include/hw/xen/xen-bus.h +1 -0
@@ -122,6 +122,7 @@ void xen_device_copy_grant_refs(XenDevice *xendev, bool to_domain,typedef void (*XenEventHandler )(void *opaque);XenEventChannel *xen_device_bind_event_channel (XenDevice *xendev, +AioContext *ctx,unsigned int port,XenEventHandler handler,void *opaque,Error **errp);
Signature · 7 lines
2.20.1.2.gb21ebb6 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH 2/3] xen-bus: allow AioContext to be specified for each event channel
Quoted 15 lines · Paul Durrant
This patch adds an AioContext parameter to
xen_device_bind_event_channel()and then usesaio_set_fd_handler()to set the callback rather thanqemu_set_fd_handler().Signed-off-by Paul Durrant <[email protected]>---
hw/xen/xen-bus.c +1
-0@@ -943,6 +944,7 @@ static void xen_device_event(void *opaque) }XenEventChannel *xen_device_bind_event_channel (XenDevice *xendev, +AioContext *ctx,unsigned int port,XenEventHandler handler,void *opaque,Error **errp)
hw/xen/xen-bus.c XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, · line 975
Quoted 8 lines
hw/xen/xen-bus.c +3
-2@@ -968,8 +970,9 @@ XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, channel->handler = handler; channel->opaque = opaque; -qemu_set_fd_handler (xenevtchn_fd (channel->xeh), xen_device_event,NULL , - channel); + channel->ctx = ctx; +aio_set_fd_handler (channel->ctx,xenevtchn_fd (channel->xeh), false, + xen_device_event,NULL ,NULL , channel);
I wonder if the 'is_external' parameter of aio_set_fd_handler shoud be true' here, instead. That flag seems to be used when making a snapshot of a blockdev, for example.
That was introduced by:
dca21ef23ba48f6f1428c59f295a857e5dc203c8^..c07bc2c1658fffeee08eb46402b2f66d55b07586
What do you think?
Signature · 6 lines
Anthony PERARD _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [Qemu-devel] [PATCH 2/3] xen-bus: allow AioContext to be specified for each event channel
In reply to Paul DurrantQuoted 15 lines · Paul Durrant
This patch adds an AioContext parameter to
xen_device_bind_event_channel()and then usesaio_set_fd_handler()to set the callback rather thanqemu_set_fd_handler().Signed-off-by Paul Durrant <[email protected]>---
hw/xen/xen-bus.c +1
-0@@ -943,6 +944,7 @@ static void xen_device_event(void *opaque) }XenEventChannel *xen_device_bind_event_channel (XenDevice *xendev, +AioContext *ctx,unsigned int port,XenEventHandler handler,void *opaque,Error **errp)
hw/xen/xen-bus.c XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, · line 975
Quoted 8 lines
hw/xen/xen-bus.c +3
-2@@ -968,8 +970,9 @@ XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, channel->handler = handler; channel->opaque = opaque; -qemu_set_fd_handler (xenevtchn_fd (channel->xeh), xen_device_event,NULL , - channel); + channel->ctx = ctx; +aio_set_fd_handler (channel->ctx,xenevtchn_fd (channel->xeh), false, + xen_device_event,NULL ,NULL , channel);
I wonder if the 'is_external' parameter of aio_set_fd_handler shoud be true' here, instead. That flag seems to be used when making a snapshot of a blockdev, for example.
That was introduced by:
dca21ef23ba48f6f1428c59f295a857e5dc203c8^..c07bc2c1658fffeee08eb46402b2f66d55b07586
What do you think?
Signature · 1 line
Anthony PERARD
Re: [Qemu-devel] [PATCH 2/3] xen-bus: allow AioContext to be specified for each event channel
Quoted 38 lines
-----Original Message-----
From: Anthony PERARD [mailto:[email protected]]
Sent: 10 April 2019 13:57
To: Paul Durrant <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; Stefano Stabellini <[email protected]>; Stefan Hajnoczi <[email protected]>; Kevin Wolf <[email protected]>; Max Reitz <[email protected]>
Subject: Re: [PATCH 2/3] xen-bus: allow AioContext to be specified for each event channelQuoted 15 lines · Paul Durrant
This patch adds an AioContext parameter to
xen_device_bind_event_channel()and then usesaio_set_fd_handler()to set the callback rather thanqemu_set_fd_handler().Signed-off-by Paul Durrant <[email protected]>---
hw/xen/xen-bus.c +1
-0@@ -943,6 +944,7 @@ static void xen_device_event(void *opaque) }XenEventChannel *xen_device_bind_event_channel (XenDevice *xendev, +AioContext *ctx,unsigned int port,XenEventHandler handler,void *opaque,Error **errp)hw/xen/xen-bus.c XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, · line 975
Quoted 8 lines
hw/xen/xen-bus.c +3
-2@@ -968,8 +970,9 @@ XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, channel->handler = handler; channel->opaque = opaque; -qemu_set_fd_handler (xenevtchn_fd (channel->xeh), xen_device_event,NULL , - channel); + channel->ctx = ctx; +aio_set_fd_handler (channel->ctx,xenevtchn_fd (channel->xeh), false, + xen_device_event,NULL ,NULL , channel);I wonder if the
'is_external' parameter of aio_set_fd_handler shoud betrue' here, instead. That flag seems to be used when making a snapshot of a blockdev, for example.
Quoted 4 lines
That was introduced by:
dca21ef23ba48f6f1428c59f295a857e5dc203c8^..c07bc2c1658fffeee08eb46402b2f66d55b07586What do you think?
Interesting. I admit I was merely transcribing what qemu_set_fd_handler() passes without really looking into the values. Looking at the arguments that virtio-blk passes to aio_set_event_notifier() though, and what 'is_external' means, it would appear that setting it to true is probably the right thing to do. Do you want me to send a v2 of the series or can you fix it up?
Cheers,
Paul
Quoted 4 lines
--
Anthony PERARD
Anthony PERARD
Reviewed-byRe: [Qemu-devel] [PATCH 2/3] xen-bus: allow AioContext to be specified for each event channel
Quoted 38 lines · Paul Durrant
Quoted 38 lines
-----Original Message-----
From: Anthony PERARD [mailto:[email protected]]
Sent: 10 April 2019 13:57
To: Paul Durrant <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; Stefano Stabellini <[email protected]>; Stefan Hajnoczi <[email protected]>; Kevin Wolf <[email protected]>; Max Reitz <[email protected]>
Subject: Re: [PATCH 2/3] xen-bus: allow AioContext to be specified for each event channelQuoted 15 lines · Paul Durrant
This patch adds an AioContext parameter to
xen_device_bind_event_channel()and then usesaio_set_fd_handler()to set the callback rather thanqemu_set_fd_handler().Signed-off-by Paul Durrant <[email protected]>---
hw/xen/xen-bus.c +1
-0@@ -943,6 +944,7 @@ static void xen_device_event(void *opaque) }XenEventChannel *xen_device_bind_event_channel (XenDevice *xendev, +AioContext *ctx,unsigned int port,XenEventHandler handler,void *opaque,Error **errp)hw/xen/xen-bus.c XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, · line 975
Quoted 8 lines
hw/xen/xen-bus.c +3
-2@@ -968,8 +970,9 @@ XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, channel->handler = handler; channel->opaque = opaque; -qemu_set_fd_handler (xenevtchn_fd (channel->xeh), xen_device_event,NULL , - channel); + channel->ctx = ctx; +aio_set_fd_handler (channel->ctx,xenevtchn_fd (channel->xeh), false, + xen_device_event,NULL ,NULL , channel);I wonder if the
'is_external' parameter of aio_set_fd_handler shoud betrue' here, instead. That flag seems to be used when making a snapshot of a blockdev, for example.
Quoted 6 lines
Quoted 4 lines
That was introduced by:
dca21ef23ba48f6f1428c59f295a857e5dc203c8^..c07bc2c1658fffeee08eb46402b2f66d55b07586What do you think?
Interesting. I admit I was merely transcribing what
qemu_set_fd_handler()passes without really looking into the values. Looking at the arguments that virtio-blk passes toaio_set_event_notifier()though, and what 'is_external' means, it would appear that setting it to true is probably the right thing to do. Do you want me to send a v2 of the series or can you fix it up?
I'll fix that up.
Thanks,
Signature · 1 line
Anthony PERARD
Re: [Qemu-devel] [PATCH 2/3] xen-bus: allow AioContext to be specified for each event channel
Quoted 53 lines
-----Original Message-----
From: Anthony PERARD [mailto:[email protected]]
Sent: 10 April 2019 16:23
To: Paul Durrant <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; Stefano Stabellini <[email protected]>; Stefan Hajnoczi <[email protected]>; Kevin Wolf <[email protected]>; Max Reitz <[email protected]>
Subject: Re: [PATCH 2/3] xen-bus: allow AioContext to be specified for each event channelQuoted 5 lines · Paul Durrant
Quoted 5 lines
-----Original Message-----
From: Anthony PERARD [mailto:[email protected]]
Sent: 10 April 2019 13:57
To: Paul Durrant <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; StefanoStabellini
Quoted 1 line
Quoted 1 line
<[email protected]>; Stefan Hajnoczi <[email protected]>; Kevin Wolf <[email protected]>;
Max
Quoted 35 lines
Quoted 35 lines
Reitz <[email protected]>
Subject: Re: [PATCH 2/3] xen-bus: allow AioContext to be specified for each event channelQuoted 15 lines · Paul Durrant
This patch adds an AioContext parameter to
xen_device_bind_event_channel()and then usesaio_set_fd_handler()to set the callback rather thanqemu_set_fd_handler().Signed-off-by Paul Durrant <[email protected]>---
hw/xen/xen-bus.c +1
-0@@ -943,6 +944,7 @@ static void xen_device_event(void *opaque) }XenEventChannel *xen_device_bind_event_channel (XenDevice *xendev, +AioContext *ctx,unsigned int port,XenEventHandler handler,void *opaque,Error **errp)hw/xen/xen-bus.c XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, · line 975
Quoted 8 lines
hw/xen/xen-bus.c +3
-2@@ -968,8 +970,9 @@ XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, channel->handler = handler; channel->opaque = opaque; -qemu_set_fd_handler (xenevtchn_fd (channel->xeh), xen_device_event,NULL , - channel); + channel->ctx = ctx; +aio_set_fd_handler (channel->ctx,xenevtchn_fd (channel->xeh), false, + xen_device_event,NULL ,NULL , channel);I wonder if the
'is_external' parameter of aio_set_fd_handler shoud betrue' here, instead. That flag seems to be used when making a snapshot of a blockdev, for example.That was introduced by:
dca21ef23ba48f6f1428c59f295a857e5dc203c8^..c07bc2c1658fffeee08eb46402b2f66d55b07586
Quoted 8 lines
Quoted 3 lines
Quoted 1 line
What do you think?
Interesting. I admit I was merely transcribing what
qemu_set_fd_handler()passes without reallylooking into the values. Looking at the arguments that virtio-blk passes to
aio_set_event_notifier()though, and what 'is_external' means, it would appear that setting it to true is probably the right thing to do. Do you want me to send a v2 of the series or can you fix it up?I'll fix that up.
Thanks,
Paul
Quoted 7 lines
Reviewed-by Anthony PERARD <[email protected]>Thanks,
--
Anthony PERARD
Anthony PERARD
Reviewed-byRe: [PATCH 2/3] xen-bus: allow AioContext to be specified for each event channel
In reply to Paul DurrantQuoted 38 lines · Paul Durrant
Quoted 38 lines
-----Original Message-----
From: Anthony PERARD [mailto:[email protected]]
Sent: 10 April 2019 13:57
To: Paul Durrant <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; Stefano Stabellini <[email protected]>; Stefan Hajnoczi <[email protected]>; Kevin Wolf <[email protected]>; Max Reitz <[email protected]>
Subject: Re: [PATCH 2/3] xen-bus: allow AioContext to be specified for each event channelQuoted 15 lines · Paul Durrant
This patch adds an AioContext parameter to
xen_device_bind_event_channel()and then usesaio_set_fd_handler()to set the callback rather thanqemu_set_fd_handler().Signed-off-by Paul Durrant <[email protected]>---
hw/xen/xen-bus.c +1
-0@@ -943,6 +944,7 @@ static void xen_device_event(void *opaque) }XenEventChannel *xen_device_bind_event_channel (XenDevice *xendev, +AioContext *ctx,unsigned int port,XenEventHandler handler,void *opaque,Error **errp)hw/xen/xen-bus.c XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, · line 975
Quoted 8 lines
hw/xen/xen-bus.c +3
-2@@ -968,8 +970,9 @@ XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, channel->handler = handler; channel->opaque = opaque; -qemu_set_fd_handler (xenevtchn_fd (channel->xeh), xen_device_event,NULL , - channel); + channel->ctx = ctx; +aio_set_fd_handler (channel->ctx,xenevtchn_fd (channel->xeh), false, + xen_device_event,NULL ,NULL , channel);I wonder if the
'is_external' parameter of aio_set_fd_handler shoud betrue' here, instead. That flag seems to be used when making a snapshot of a blockdev, for example.
Quoted 6 lines
Quoted 4 lines
That was introduced by:
dca21ef23ba48f6f1428c59f295a857e5dc203c8^..c07bc2c1658fffeee08eb46402b2f66d55b07586What do you think?
Interesting. I admit I was merely transcribing what
qemu_set_fd_handler()passes without really looking into the values. Looking at the arguments that virtio-blk passes toaio_set_event_notifier()though, and what 'is_external' means, it would appear that setting it to true is probably the right thing to do. Do you want me to send a v2 of the series or can you fix it up?
I'll fix that up.
Thanks,
Signature · 6 lines
Anthony PERARD _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [Qemu-block] [PATCH 2/3] xen-bus: allow AioContext to be specified for each event channel
In reply to Paul DurrantQuoted 38 lines · Paul Durrant
Quoted 38 lines
-----Original Message-----
From: Anthony PERARD [mailto:[email protected]]
Sent: 10 April 2019 13:57
To: Paul Durrant <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; Stefano Stabellini <[email protected]>; Stefan Hajnoczi <[email protected]>; Kevin Wolf <[email protected]>; Max Reitz <[email protected]>
Subject: Re: [PATCH 2/3] xen-bus: allow AioContext to be specified for each event channelQuoted 15 lines · Paul Durrant
This patch adds an AioContext parameter to
xen_device_bind_event_channel()and then usesaio_set_fd_handler()to set the callback rather thanqemu_set_fd_handler().Signed-off-by Paul Durrant <[email protected]>---
hw/xen/xen-bus.c +1
-0@@ -943,6 +944,7 @@ static void xen_device_event(void *opaque) }XenEventChannel *xen_device_bind_event_channel (XenDevice *xendev, +AioContext *ctx,unsigned int port,XenEventHandler handler,void *opaque,Error **errp)hw/xen/xen-bus.c XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, · line 975
Quoted 8 lines
hw/xen/xen-bus.c +3
-2@@ -968,8 +970,9 @@ XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, channel->handler = handler; channel->opaque = opaque; -qemu_set_fd_handler (xenevtchn_fd (channel->xeh), xen_device_event,NULL , - channel); + channel->ctx = ctx; +aio_set_fd_handler (channel->ctx,xenevtchn_fd (channel->xeh), false, + xen_device_event,NULL ,NULL , channel);I wonder if the
'is_external' parameter of aio_set_fd_handler shoud betrue' here, instead. That flag seems to be used when making a snapshot of a blockdev, for example.
Quoted 6 lines
Quoted 4 lines
That was introduced by:
dca21ef23ba48f6f1428c59f295a857e5dc203c8^..c07bc2c1658fffeee08eb46402b2f66d55b07586What do you think?
Interesting. I admit I was merely transcribing what
qemu_set_fd_handler()passes without really looking into the values. Looking at the arguments that virtio-blk passes toaio_set_event_notifier()though, and what 'is_external' means, it would appear that setting it to true is probably the right thing to do. Do you want me to send a v2 of the series or can you fix it up?
Hi,
Handlers are invoked by the aio_poll() event loop. Some handlers are considered "external" in the sense that they submit new I/O requests from the guest or outside world. Others are considered "internal" in the sense that they are part of the block layer and not an entry point into the block layer.
There are points where the block layer wants to run the event loop but new requests must not be submitted. In this case aio_disable_external() will be called so that "external" handlers are not processed.
For example, see virtio's virtio_queue_aio_set_host_notifier_handler(). This is the virtqueue kick ioeventfd and it shouldn't be processed when aio_disable_external() has been called.
Stefan
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [Qemu-devel] [Qemu-block] [PATCH 2/3] xen-bus: allow AioContext to be specified for each event channel
In reply to Paul DurrantQuoted 38 lines · Paul Durrant
Quoted 38 lines
-----Original Message-----
From: Anthony PERARD [mailto:[email protected]]
Sent: 10 April 2019 13:57
To: Paul Durrant <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; Stefano Stabellini <[email protected]>; Stefan Hajnoczi <[email protected]>; Kevin Wolf <[email protected]>; Max Reitz <[email protected]>
Subject: Re: [PATCH 2/3] xen-bus: allow AioContext to be specified for each event channelQuoted 15 lines · Paul Durrant
This patch adds an AioContext parameter to
xen_device_bind_event_channel()and then usesaio_set_fd_handler()to set the callback rather thanqemu_set_fd_handler().Signed-off-by Paul Durrant <[email protected]>---
hw/xen/xen-bus.c +1
-0@@ -943,6 +944,7 @@ static void xen_device_event(void *opaque) }XenEventChannel *xen_device_bind_event_channel (XenDevice *xendev, +AioContext *ctx,unsigned int port,XenEventHandler handler,void *opaque,Error **errp)hw/xen/xen-bus.c XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, · line 975
Quoted 8 lines
hw/xen/xen-bus.c +3
-2@@ -968,8 +970,9 @@ XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, channel->handler = handler; channel->opaque = opaque; -qemu_set_fd_handler (xenevtchn_fd (channel->xeh), xen_device_event,NULL , - channel); + channel->ctx = ctx; +aio_set_fd_handler (channel->ctx,xenevtchn_fd (channel->xeh), false, + xen_device_event,NULL ,NULL , channel);I wonder if the
'is_external' parameter of aio_set_fd_handler shoud betrue' here, instead. That flag seems to be used when making a snapshot of a blockdev, for example.
Quoted 6 lines
Quoted 4 lines
That was introduced by:
dca21ef23ba48f6f1428c59f295a857e5dc203c8^..c07bc2c1658fffeee08eb46402b2f66d55b07586What do you think?
Interesting. I admit I was merely transcribing what
qemu_set_fd_handler()passes without really looking into the values. Looking at the arguments that virtio-blk passes toaio_set_event_notifier()though, and what 'is_external' means, it would appear that setting it to true is probably the right thing to do. Do you want me to send a v2 of the series or can you fix it up?
Hi,
Handlers are invoked by the aio_poll() event loop. Some handlers are considered "external" in the sense that they submit new I/O requests from the guest or outside world. Others are considered "internal" in the sense that they are part of the block layer and not an entry point into the block layer.
There are points where the block layer wants to run the event loop but new requests must not be submitted. In this case aio_disable_external() will be called so that "external" handlers are not processed.
For example, see virtio's virtio_queue_aio_set_host_notifier_handler(). This is the virtqueue kick ioeventfd and it shouldn't be processed when aio_disable_external() has been called.
Stefan
Re: [Qemu-devel] [Qemu-block] [PATCH 2/3] xen-bus: allow AioContext to be specified for each event channel
Quoted 1 line
-----Original Message-----
[snip]
Quoted 21 lines
Quoted 10 lines
Quoted 8 lines
I wonder if the
'is_external' parameter of aio_set_fd_handler shoud betrue' here, instead. That flag seems to be used when making a snapshot of a blockdev, for example.That was introduced by:
dca21ef23ba48f6f1428c59f295a857e5dc203c8^..c07bc2c1658fffeee08eb46402b2f66d55b07586What do you think?
Interesting. I admit I was merely transcribing what
qemu_set_fd_handler()passes without reallylooking into the values. Looking at the arguments that virtio-blk passes to
aio_set_event_notifier()though, and what 'is_external' means, it would appear that setting it to true is probably the right thing to do. Do you want me to send a v2 of the series or can you fix it up?Hi,
Handlers are invoked by theaio_poll()event loop. Some handlers are considered "external" in the sense that they submit new I/O requests from the guest or outside world. Others are considered "internal" in the sense that they are part of the block layer and not an entry point into the block layer.
Quoted 8 lines
There are points where the block layer wants to run the event loop but new requests must not be submitted. In this case
aio_disable_external()will be called so that "external" handlers are not processed.For example, see virtio's
virtio_queue_aio_set_host_notifier_handler(). This is the virtqueue kick ioeventfd and it shouldn't be processed whenaio_disable_external()has been called.
Thanks for the explanation Stefan. Xen event channels/shared rings should indeed be considered as external sources.
Cheers,
Paul
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [Qemu-devel] [Qemu-block] [PATCH 2/3] xen-bus: allow AioContext to be specified for each event channel
In reply to Stefan HajnocziQuoted 1 line
-----Original Message-----
[snip]
Quoted 21 lines
Quoted 10 lines
Quoted 8 lines
I wonder if the
'is_external' parameter of aio_set_fd_handler shoud betrue' here, instead. That flag seems to be used when making a snapshot of a blockdev, for example.That was introduced by:
dca21ef23ba48f6f1428c59f295a857e5dc203c8^..c07bc2c1658fffeee08eb46402b2f66d55b07586What do you think?
Interesting. I admit I was merely transcribing what
qemu_set_fd_handler()passes without reallylooking into the values. Looking at the arguments that virtio-blk passes to
aio_set_event_notifier()though, and what 'is_external' means, it would appear that setting it to true is probably the right thing to do. Do you want me to send a v2 of the series or can you fix it up?Hi,
Handlers are invoked by theaio_poll()event loop. Some handlers are considered "external" in the sense that they submit new I/O requests from the guest or outside world. Others are considered "internal" in the sense that they are part of the block layer and not an entry point into the block layer.
Quoted 8 lines
There are points where the block layer wants to run the event loop but new requests must not be submitted. In this case
aio_disable_external()will be called so that "external" handlers are not processed.For example, see virtio's
virtio_queue_aio_set_host_notifier_handler(). This is the virtqueue kick ioeventfd and it shouldn't be processed whenaio_disable_external()has been called.
Thanks for the explanation Stefan. Xen event channels/shared rings should indeed be considered as external sources.
Cheers,
Paul
Re: [PATCH 2/3] xen-bus: allow AioContext to be specified for each event channel
In reply to Anthony PERARDQuoted 38 lines
-----Original Message-----
From: Anthony PERARD [mailto:[email protected]]
Sent: 10 April 2019 13:57
To: Paul Durrant <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; Stefano Stabellini <[email protected]>; Stefan Hajnoczi <[email protected]>; Kevin Wolf <[email protected]>; Max Reitz <[email protected]>
Subject: Re: [PATCH 2/3] xen-bus: allow AioContext to be specified for each event channelQuoted 15 lines · Paul Durrant
This patch adds an AioContext parameter to
xen_device_bind_event_channel()and then usesaio_set_fd_handler()to set the callback rather thanqemu_set_fd_handler().Signed-off-by Paul Durrant <[email protected]>---
hw/xen/xen-bus.c +1
-0@@ -943,6 +944,7 @@ static void xen_device_event(void *opaque) }XenEventChannel *xen_device_bind_event_channel (XenDevice *xendev, +AioContext *ctx,unsigned int port,XenEventHandler handler,void *opaque,Error **errp)hw/xen/xen-bus.c XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, · line 975
Quoted 8 lines
hw/xen/xen-bus.c +3
-2@@ -968,8 +970,9 @@ XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, channel->handler = handler; channel->opaque = opaque; -qemu_set_fd_handler (xenevtchn_fd (channel->xeh), xen_device_event,NULL , - channel); + channel->ctx = ctx; +aio_set_fd_handler (channel->ctx,xenevtchn_fd (channel->xeh), false, + xen_device_event,NULL ,NULL , channel);I wonder if the
'is_external' parameter of aio_set_fd_handler shoud betrue' here, instead. That flag seems to be used when making a snapshot of a blockdev, for example.
Quoted 4 lines
That was introduced by:
dca21ef23ba48f6f1428c59f295a857e5dc203c8^..c07bc2c1658fffeee08eb46402b2f66d55b07586What do you think?
Interesting. I admit I was merely transcribing what qemu_set_fd_handler() passes without really looking into the values. Looking at the arguments that virtio-blk passes to aio_set_event_notifier() though, and what 'is_external' means, it would appear that setting it to true is probably the right thing to do. Do you want me to send a v2 of the series or can you fix it up?
Cheers,
Paul
Quoted 4 lines
--
Anthony PERARD
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
PATCH 3/3 xen-bus / xen-block: add support for event channel polling
In reply to Paul DurrantThis patch introduces a poll callback for event channel fd-s and uses this to invoke the channel callback function.
To properly support polling, it is necessary for the event channel callback function to return a boolean saying whether it has done any useful work or not. Thus xen_block_dataplane_event() is modified to directly invoke xen_block_handle_requests() and the latter only returns true if it actually processes any requests. This also means that the call to qemu_bh_schedule() is moved into xen_block_complete_aio(), which is more intuitive since the only reason for doing a deferred poll of the shared ring should be because there were previously insufficient resources to fully complete a previous poll.
---
Cc: Stefan Hajnoczi <[email protected]>
Cc: Stefano Stabellini <[email protected]>
Cc: Anthony Perard <[email protected]>
Cc: Kevin Wolf <[email protected]>
Cc: Max Reitz <[email protected]>
--- hw/block/dataplane/xen-block.c | 17 +++++++++-------- hw/xen/xen-bus.c | 11 +++++++++-- include/hw/xen/xen-bus.h | 2 +- 3 files changed, 19 insertions(+), 11 deletions(-)
hw/block/dataplane/xen-block.c +9 -8
@@ -317,7 +317,9 @@ static void xen_block_complete_aio(void *opaque, int ret) }xen_block_release_request (request); -qemu_bh_schedule (dataplane->bh); +if (dataplane->more_work) { +qemu_bh_schedule (dataplane->bh); + } done:aio_context_release (dataplane->ctx);
@@ -514,12 +516,13 @@ static int xen_block_get_request(XenBlockDataPlane *dataplane,*/ #define IO_PLUG_THRESHOLD 1 -static void xen_block_handle_requests (XenBlockDataPlane *dataplane) +static bool xen_block_handle_requests (XenBlockDataPlane *dataplane) {RING_IDX rc, rp;XenBlockRequest *request;int inflight_atstart = dataplane->requests_inflight;int batched =0 ; +bool done_something = false; dataplane->more_work =0 ;
@@ -551,6 +554,7 @@ static void xen_block_handle_requests(XenBlockDataPlane *dataplane) }xen_block_get_request (dataplane, request, rc); dataplane->rings.common.req_cons = ++rc; + done_something = true;/* parse them */ if (xen_block_parse_request (request) !=0 ) {
@@ -602,10 +606,7 @@ static void xen_block_handle_requests(XenBlockDataPlane *dataplane)blk_io_unplug (dataplane->blk); } -if (dataplane->more_work && - dataplane->requests_inflight < dataplane->max_requests) { -qemu_bh_schedule (dataplane->bh); - } +return done_something; }static void xen_block_dataplane_bh (void *opaque)
@@ -617,11 +618,11 @@ static void xen_block_dataplane_bh(void *opaque)aio_context_release (dataplane->ctx); } -static void xen_block_dataplane_event (void *opaque) +static bool xen_block_dataplane_event (void *opaque) {XenBlockDataPlane *dataplane = opaque; -qemu_bh_schedule (dataplane->bh); +return xen_block_handle_requests (dataplane); } XenBlockDataPlane *xen_block_dataplane_create(XenDevice *xendev,
hw/xen/xen-bus.c +9 -2
@@ -931,13 +931,20 @@ struct XenEventChannel {void *opaque; }; +static bool xen_device_poll (void *opaque) +{ +XenEventChannel *channel = opaque; + +return channel->handler (channel->opaque); +} +static void xen_device_event (void *opaque) {XenEventChannel *channel = opaque;unsigned long port =xenevtchn_pending (channel->xeh);if (port == channel->local_port) { - channel->handler (channel->opaque); +xen_device_poll (channel);xenevtchn_unmask (channel->xeh, port); }
@@ -972,7 +979,7 @@ XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, channel->ctx = ctx;aio_set_fd_handler (channel->ctx,xenevtchn_fd (channel->xeh), false, - xen_device_event,NULL ,NULL , channel); + xen_device_event,NULL , xen_device_poll, channel);QLIST_INSERT_HEAD (&xendev->event_channels, channel, list);
include/hw/xen/xen-bus.h +1 -1
@@ -119,7 +119,7 @@ void xen_device_copy_grant_refs(XenDevice *xendev, bool to_domain,XenDeviceGrantCopySegment segs[], unsigned int nr_segs, Error **errp); -typedef void (*XenEventHandler )(void *opaque); +typedef bool (*XenEventHandler )(void *opaque); XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, AioContext *ctx,
Signature · 7 lines
2.20.1.2.gb21ebb6 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
PATCH 3/3 xen-bus / xen-block: add support for event channel polling
In reply to Paul DurrantThis patch introduces a poll callback for event channel fd-s and uses this to invoke the channel callback function.
To properly support polling, it is necessary for the event channel callback function to return a boolean saying whether it has done any useful work or not. Thus xen_block_dataplane_event() is modified to directly invoke xen_block_handle_requests() and the latter only returns true if it actually processes any requests. This also means that the call to qemu_bh_schedule() is moved into xen_block_complete_aio(), which is more intuitive since the only reason for doing a deferred poll of the shared ring should be because there were previously insufficient resources to fully complete a previous poll.
---
Cc: Stefan Hajnoczi <[email protected]>
Cc: Stefano Stabellini <[email protected]>
Cc: Anthony Perard <[email protected]>
Cc: Kevin Wolf <[email protected]>
Cc: Max Reitz <[email protected]>
--- hw/block/dataplane/xen-block.c | 17 +++++++++-------- hw/xen/xen-bus.c | 11 +++++++++-- include/hw/xen/xen-bus.h | 2 +- 3 files changed, 19 insertions(+), 11 deletions(-)
hw/block/dataplane/xen-block.c +9 -8
@@ -317,7 +317,9 @@ static void xen_block_complete_aio(void *opaque, int ret) }xen_block_release_request (request); -qemu_bh_schedule (dataplane->bh); +if (dataplane->more_work) { +qemu_bh_schedule (dataplane->bh); + } done:aio_context_release (dataplane->ctx);
@@ -514,12 +516,13 @@ static int xen_block_get_request(XenBlockDataPlane *dataplane,*/ #define IO_PLUG_THRESHOLD 1 -static void xen_block_handle_requests (XenBlockDataPlane *dataplane) +static bool xen_block_handle_requests (XenBlockDataPlane *dataplane) {RING_IDX rc, rp;XenBlockRequest *request;int inflight_atstart = dataplane->requests_inflight;int batched =0 ; +bool done_something = false; dataplane->more_work =0 ;
@@ -551,6 +554,7 @@ static void xen_block_handle_requests(XenBlockDataPlane *dataplane) }xen_block_get_request (dataplane, request, rc); dataplane->rings.common.req_cons = ++rc; + done_something = true;/* parse them */ if (xen_block_parse_request (request) !=0 ) {
@@ -602,10 +606,7 @@ static void xen_block_handle_requests(XenBlockDataPlane *dataplane)blk_io_unplug (dataplane->blk); } -if (dataplane->more_work && - dataplane->requests_inflight < dataplane->max_requests) { -qemu_bh_schedule (dataplane->bh); - } +return done_something; }static void xen_block_dataplane_bh (void *opaque)
@@ -617,11 +618,11 @@ static void xen_block_dataplane_bh(void *opaque)aio_context_release (dataplane->ctx); } -static void xen_block_dataplane_event (void *opaque) +static bool xen_block_dataplane_event (void *opaque) {XenBlockDataPlane *dataplane = opaque; -qemu_bh_schedule (dataplane->bh); +return xen_block_handle_requests (dataplane); } XenBlockDataPlane *xen_block_dataplane_create(XenDevice *xendev,
hw/xen/xen-bus.c +9 -2
@@ -931,13 +931,20 @@ struct XenEventChannel {void *opaque; }; +static bool xen_device_poll (void *opaque) +{ +XenEventChannel *channel = opaque; + +return channel->handler (channel->opaque); +} +static void xen_device_event (void *opaque) {XenEventChannel *channel = opaque;unsigned long port =xenevtchn_pending (channel->xeh);if (port == channel->local_port) { - channel->handler (channel->opaque); +xen_device_poll (channel);xenevtchn_unmask (channel->xeh, port); }
@@ -972,7 +979,7 @@ XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, channel->ctx = ctx;aio_set_fd_handler (channel->ctx,xenevtchn_fd (channel->xeh), false, - xen_device_event,NULL ,NULL , channel); + xen_device_event,NULL , xen_device_poll, channel);QLIST_INSERT_HEAD (&xendev->event_channels, channel, list);
include/hw/xen/xen-bus.h +1 -1
@@ -119,7 +119,7 @@ void xen_device_copy_grant_refs(XenDevice *xendev, bool to_domain,XenDeviceGrantCopySegment segs[], unsigned int nr_segs, Error **errp); -typedef void (*XenEventHandler )(void *opaque); +typedef bool (*XenEventHandler )(void *opaque); XenEventChannel *xen_device_bind_event_channel(XenDevice *xendev, AioContext *ctx,
Signature · 1 line
2.20.1.2.gb21ebb6
Re: [Qemu-devel] [PATCH 3/3] xen-bus / xen-block: add support for event channel polling
Quoted 13 lines · Paul Durrant
This patch introduces a poll callback for event channel fd-s and uses this to invoke the channel callback function.
To properly support polling, it is necessary for the event channel callback function to return a boolean saying whether it has done any useful work or not. Thus
xen_block_dataplane_event()is modified to directly invokexen_block_handle_requests()and the latter only returns true if it actually processes any requests. This also means that the call toqemu_bh_schedule()is moved intoxen_block_complete_aio(), which is more intuitive since the only reason for doing a deferred poll of the shared ring should be because there were previously insufficient resources to fully complete a previous poll.
Quoted 2 lines
Signed-off-by Paul Durrant <[email protected]>---
Thanks,
Signature · 6 lines
Anthony PERARD _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel