| Message ID | 1501119055-4060-3-git-send-email-mdroth@linux.vnet.ibm.com |
|---|---|
| State | New |
| Headers | show |
On Wed, 26 Jul 2017 20:30:54 -0500 Michael Roth <mdroth@linux.vnet.ibm.com> wrote: > This reverts commit abed886ec60cf239a03515cf0b30fb11fa964c44. > > This patch originally addressed an issue where a DEVICE_DELETED > event could be emitted (in device_unparent()) before a Device's > QemuOpts were cleaned up (in device_finalize()), leading to a > "duplicate ID" error if management attempted to immediately add > a device with the same ID in response to the DEVICE_DELETED event. > > An alternative will be implemented in a subsequent patch where we > defer the DEVICE_DELETED event until device_finalize(), which would > also prevent the race, so we revert the original fix in preparation. > Do you really need to revert abed886ec60cf239a03515cf0b30fb11fa964c44 ? IIUC, the purpose of this series is to fix/workaround the VFIO cleanup issue. Even if it also relates to when we emit DEVICE_DELETED, it isn't exactly the same problem here, is it ? > Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com> > --- > hw/core/qdev.c | 4 +--- > 1 file changed, 1 insertion(+), 3 deletions(-) > > diff --git a/hw/core/qdev.c b/hw/core/qdev.c > index a64b35c..08c4061 100644 > --- a/hw/core/qdev.c > +++ b/hw/core/qdev.c > @@ -1067,6 +1067,7 @@ static void device_finalize(Object *obj) > NamedGPIOList *ngl, *next; > > DeviceState *dev = DEVICE(obj); > + qemu_opts_del(dev->opts); > > QLIST_FOREACH_SAFE(ngl, &dev->gpios, node, next) { > QLIST_REMOVE(ngl, node); > @@ -1116,9 +1117,6 @@ static void device_unparent(Object *obj) > g_free(dev->canonical_path); > dev->canonical_path = NULL; > } > - > - qemu_opts_del(dev->opts); > - dev->opts = NULL; > } > > static void device_class_init(ObjectClass *class, void *data)
Quoting Greg Kurz (2017-07-31 10:51:39) > On Wed, 26 Jul 2017 20:30:54 -0500 > Michael Roth <mdroth@linux.vnet.ibm.com> wrote: > > > This reverts commit abed886ec60cf239a03515cf0b30fb11fa964c44. > > > > This patch originally addressed an issue where a DEVICE_DELETED > > event could be emitted (in device_unparent()) before a Device's > > QemuOpts were cleaned up (in device_finalize()), leading to a > > "duplicate ID" error if management attempted to immediately add > > a device with the same ID in response to the DEVICE_DELETED event. > > > > An alternative will be implemented in a subsequent patch where we > > defer the DEVICE_DELETED event until device_finalize(), which would > > also prevent the race, so we revert the original fix in preparation. > > > > Do you really need to revert abed886ec60cf239a03515cf0b30fb11fa964c44 ? > > IIUC, the purpose of this series is to fix/workaround the VFIO cleanup issue. > Even if it also relates to when we emit DEVICE_DELETED, it isn't exactly the > same problem here, is it ? That's true, but since the implementation of the VFIO fix requires moving qemu_opts_del() to back to finalize(), and since that effectively reverts the change made in abed88, I figured it was better to revert the patch directly as opposed as a side-effect of patch 3. > > > Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com> > > --- > > hw/core/qdev.c | 4 +--- > > 1 file changed, 1 insertion(+), 3 deletions(-) > > > > diff --git a/hw/core/qdev.c b/hw/core/qdev.c > > index a64b35c..08c4061 100644 > > --- a/hw/core/qdev.c > > +++ b/hw/core/qdev.c > > @@ -1067,6 +1067,7 @@ static void device_finalize(Object *obj) > > NamedGPIOList *ngl, *next; > > > > DeviceState *dev = DEVICE(obj); > > + qemu_opts_del(dev->opts); > > > > QLIST_FOREACH_SAFE(ngl, &dev->gpios, node, next) { > > QLIST_REMOVE(ngl, node); > > @@ -1116,9 +1117,6 @@ static void device_unparent(Object *obj) > > g_free(dev->canonical_path); > > dev->canonical_path = NULL; > > } > > - > > - qemu_opts_del(dev->opts); > > - dev->opts = NULL; > > } > > > > static void device_class_init(ObjectClass *class, void *data) >
On Mon, 31 Jul 2017 11:39:13 -0500 Michael Roth <mdroth@linux.vnet.ibm.com> wrote: > Quoting Greg Kurz (2017-07-31 10:51:39) > > On Wed, 26 Jul 2017 20:30:54 -0500 > > Michael Roth <mdroth@linux.vnet.ibm.com> wrote: > > > > > This reverts commit abed886ec60cf239a03515cf0b30fb11fa964c44. > > > > > > This patch originally addressed an issue where a DEVICE_DELETED > > > event could be emitted (in device_unparent()) before a Device's > > > QemuOpts were cleaned up (in device_finalize()), leading to a > > > "duplicate ID" error if management attempted to immediately add > > > a device with the same ID in response to the DEVICE_DELETED event. > > > > > > An alternative will be implemented in a subsequent patch where we > > > defer the DEVICE_DELETED event until device_finalize(), which would > > > also prevent the race, so we revert the original fix in preparation. > > > > > > > Do you really need to revert abed886ec60cf239a03515cf0b30fb11fa964c44 ? > > > > IIUC, the purpose of this series is to fix/workaround the VFIO cleanup issue. > > Even if it also relates to when we emit DEVICE_DELETED, it isn't exactly the > > same problem here, is it ? > > That's true, but since the implementation of the VFIO fix requires > moving qemu_opts_del() to back to finalize(), and since that > effectively reverts the change made in abed88, I figured it was > better to revert the patch directly as opposed as a side-effect > of patch 3. > Fair enough. Reviewed-by: Greg Kurz <groug@kaod.org> > > > > > Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com> > > > --- > > > hw/core/qdev.c | 4 +--- > > > 1 file changed, 1 insertion(+), 3 deletions(-) > > > > > > diff --git a/hw/core/qdev.c b/hw/core/qdev.c > > > index a64b35c..08c4061 100644 > > > --- a/hw/core/qdev.c > > > +++ b/hw/core/qdev.c > > > @@ -1067,6 +1067,7 @@ static void device_finalize(Object *obj) > > > NamedGPIOList *ngl, *next; > > > > > > DeviceState *dev = DEVICE(obj); > > > + qemu_opts_del(dev->opts); > > > > > > QLIST_FOREACH_SAFE(ngl, &dev->gpios, node, next) { > > > QLIST_REMOVE(ngl, node); > > > @@ -1116,9 +1117,6 @@ static void device_unparent(Object *obj) > > > g_free(dev->canonical_path); > > > dev->canonical_path = NULL; > > > } > > > - > > > - qemu_opts_del(dev->opts); > > > - dev->opts = NULL; > > > } > > > > > > static void device_class_init(ObjectClass *class, void *data) > > >
diff --git a/hw/core/qdev.c b/hw/core/qdev.c index a64b35c..08c4061 100644 --- a/hw/core/qdev.c +++ b/hw/core/qdev.c @@ -1067,6 +1067,7 @@ static void device_finalize(Object *obj) NamedGPIOList *ngl, *next; DeviceState *dev = DEVICE(obj); + qemu_opts_del(dev->opts); QLIST_FOREACH_SAFE(ngl, &dev->gpios, node, next) { QLIST_REMOVE(ngl, node); @@ -1116,9 +1117,6 @@ static void device_unparent(Object *obj) g_free(dev->canonical_path); dev->canonical_path = NULL; } - - qemu_opts_del(dev->opts); - dev->opts = NULL; } static void device_class_init(ObjectClass *class, void *data)
This reverts commit abed886ec60cf239a03515cf0b30fb11fa964c44. This patch originally addressed an issue where a DEVICE_DELETED event could be emitted (in device_unparent()) before a Device's QemuOpts were cleaned up (in device_finalize()), leading to a "duplicate ID" error if management attempted to immediately add a device with the same ID in response to the DEVICE_DELETED event. An alternative will be implemented in a subsequent patch where we defer the DEVICE_DELETED event until device_finalize(), which would also prevent the race, so we revert the original fix in preparation. Signed-off-by: Michael Roth <mdroth@linux.vnet.ibm.com> --- hw/core/qdev.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-)