hw/arm: Use ARM_CPU_TYPE_NAME() and object_initialize_child()
Subsystems (5)
cover hw/arm: Use ARM_CPU_TYPE_NAME() and object_initialize_child()
First we use ARM_CPU_TYPE_NAME() when we should.
Then is follow up of
https://lists.gnu.org/archive/html/qemu-devel/2019-05/msg01492.html
This series looks at Eduardo suggestions from [1] and Thomas commit aff39be0ed97 to replace various object_initialize + qdev_set_parent_bus calls by sysbus_init_child_obj().
Finally, some devices are declared orphean while they have a parent, let them be together again.
[1] https://patchwork.ozlabs.org/patch/943333/#1953608
Philippe Mathieu-Daudé (6):
hw/arm: Use ARM_CPU_TYPE_NAME() macro when appropriate hw/arm: Use object_initialize_child for correct reference counting hw/arm: Use sysbus_init_child_obj for correct reference counting hw/arm/fsl-imx: Add the cpu as child of the SoC object hw/dma/xilinx_axi: Use object_initialize_child for correct ref.
counting
hw/net/xilinx_axi: Use object_initialize_child for correct ref.
counting
hw/arm/allwinner-a10.c | 3 ++- hw/arm/cubieboard.c | 3 ++- hw/arm/digic.c | 3 ++- hw/arm/exynos4_boards.c | 4 ++-- hw/arm/fsl-imx25.c | 4 +++- hw/arm/fsl-imx31.c | 4 +++- hw/arm/fsl-imx6.c | 3 ++- hw/arm/fsl-imx6ul.c | 3 ++- hw/arm/mcimx7d-sabre.c | 9 ++++----- hw/arm/mps2-tz.c | 15 +++++++-------- hw/arm/musca.c | 9 +++++---- hw/arm/xlnx-zynqmp.c | 8 ++++---- hw/dma/xilinx_axidma.c | 16 ++++++++-------- hw/net/xilinx_axienet.c | 17 ++++++++--------- 14 files changed, 54 insertions(+), 47 deletions(-)
Signature · 1 line
2.20.1
PATCH 1/6 hw/arm: Use ARM_CPU_TYPE_NAME() macro when appropriate
Commit ba1ba5cca introduce the ARM_CPU_TYPE_NAME() macro. Unify the code base by use it in all places.
--- hw/arm/allwinner-a10.c | 3 ++- hw/arm/cubieboard.c | 3 ++- hw/arm/digic.c | 3 ++- hw/arm/fsl-imx6.c | 3 ++- hw/arm/fsl-imx6ul.c | 3 ++- hw/arm/xlnx-zynqmp.c | 8 ++++---- 6 files changed, 14 insertions(+), 9 deletions(-)
hw/arm/allwinner-a10.c +2 -1
@@ -28,7 +28,8 @@ static void aw_a10_init(Object *obj)AwA10State *s =AW_A10 (obj);object_initialize_child (obj,"cpu" , &s->cpu,sizeof (s->cpu), -"cortex-a8-" TYPE_ARM_CPU , &error_abort,NULL ); +ARM_CPU_TYPE_NAME ("cortex-a8" ), + &error_abort,NULL );sysbus_init_child_obj (obj,"intc" , &s->intc,sizeof (s->intc),TYPE_AW_A10_PIC );
hw/arm/cubieboard.c +2 -1
@@ -80,7 +80,8 @@ static void cubieboard_init(MachineState *machine)static void cubieboard_machine_init (MachineClass *mc) { - mc->desc ="cubietech cubieboard" ; + mc->desc ="cubietech cubieboard (Cortex-A9)" ; + mc->default_cpu_type =ARM_CPU_TYPE_NAME ("cortex-a9" ); mc->init = cubieboard_init; mc->block_default_type =IF_IDE ; mc->units_per_default_bus =1 ;
hw/arm/digic.c +2 -1
@@ -36,7 +36,8 @@ static void digic_init(Object *obj)int i;object_initialize_child (obj,"cpu" , &s->cpu,sizeof (s->cpu), -"arm946-" TYPE_ARM_CPU , &error_abort,NULL ); +ARM_CPU_TYPE_NAME ("arm946" ), + &error_abort,NULL );for (i =0 ; i <DIGIC4_NB_TIMERS ; i++) {#define DIGIC_TIMER_NAME_MLEN 11
hw/arm/fsl-imx6.c +2 -1
@@ -40,7 +40,8 @@ static void fsl_imx6_init(Object *obj)for (i =0 ; i <MIN (smp_cpus,FSL_IMX6_NUM_CPUS ); i++) {snprintf (name,NAME_SIZE ,"cpu%d" , i);object_initialize_child (obj, name, &s->cpu[i],sizeof (s->cpu[i]), -"cortex-a9-" TYPE_ARM_CPU , &error_abort,NULL ); +ARM_CPU_TYPE_NAME ("cortex-a9" ), + &error_abort,NULL ); } sysbus_init_child_obj(obj, "a9mpcore", &s->a9mpcore, sizeof(s->a9mpcore),
hw/arm/fsl-imx6ul.c +2 -1
@@ -35,7 +35,8 @@ static void fsl_imx6ul_init(Object *obj)for (i =0 ; i <MIN (smp_cpus,FSL_IMX6UL_NUM_CPUS ); i++) {snprintf (name,NAME_SIZE ,"cpu%d" , i);object_initialize_child (obj, name, &s->cpu[i],sizeof (s->cpu[i]), -"cortex-a7-" TYPE_ARM_CPU , &error_abort,NULL ); +ARM_CPU_TYPE_NAME ("cortex-a7" ), + &error_abort,NULL ); }/*
hw/arm/xlnx-zynqmp.c +4 -4
@@ -193,8 +193,8 @@ static void xlnx_zynqmp_create_rpu(XlnxZynqMPState *s, const char *boot_cpu,object_initialize_child (OBJECT (&s->rpu_cluster),"rpu-cpu[*]" , &s->rpu_cpu[i],sizeof (s->rpu_cpu[i]), -"cortex-r5f-" TYPE_ARM_CPU , &error_abort, -NULL ); +ARM_CPU_TYPE_NAME ("cortex-r5f" ), + &error_abort,NULL ); name =object_get_canonical_path_component (OBJECT (&s->rpu_cpu[i]));if (strcmp (name, boot_cpu)) {
@@ -233,8 +233,8 @@ static void xlnx_zynqmp_init(Object *obj)for (i =0 ; i < num_apus; i++) {object_initialize_child (OBJECT (&s->apu_cluster),"apu-cpu[*]" , &s->apu_cpu[i],sizeof (s->apu_cpu[i]), -"cortex-a53-" TYPE_ARM_CPU , &error_abort, -NULL ); +ARM_CPU_TYPE_NAME ("cortex-a53" ), + &error_abort,NULL ); } sysbus_init_child_obj(obj, "gic", &s->gic, sizeof(s->gic),
Signature · 1 line
2.20.1
Re: [Qemu-devel] [PATCH 1/6] hw/arm: Use ARM_CPU_TYPE_NAME() macro when appropriate
Quoted 5 lines · Philippe Mathieu-Daudé
Commit
ba1ba5ccaintroduce theARM_CPU_TYPE_NAME()macro. Unify the code base by use it in all places.Signed-off-by Philippe Mathieu-Daudé <[email protected]>
Alistair
Quoted 109 lines
--- hw/arm/allwinner-a10.c | 3 ++- hw/arm/cubieboard.c | 3 ++- hw/arm/digic.c | 3 ++- hw/arm/fsl-imx6.c | 3 ++- hw/arm/fsl-imx6ul.c | 3 ++- hw/arm/xlnx-zynqmp.c | 8 ++++---- 6 files changed, 14 insertions(+), 9 deletions(-)hw/arm/allwinner-a10.c +2
-1@@ -28,7 +28,8 @@ static void aw_a10_init(Object *obj)AwA10State *s =AW_A10 (obj);object_initialize_child (obj,"cpu" , &s->cpu,sizeof (s->cpu), -"cortex-a8-" TYPE_ARM_CPU , &error_abort,NULL ); +ARM_CPU_TYPE_NAME ("cortex-a8" ), + &error_abort,NULL );sysbus_init_child_obj (obj,"intc" , &s->intc,sizeof (s->intc),TYPE_AW_A10_PIC );hw/arm/cubieboard.c +2
-1@@ -80,7 +80,8 @@ static void cubieboard_init(MachineState *machine)static void cubieboard_machine_init (MachineClass *mc) { - mc->desc ="cubietech cubieboard" ; + mc->desc ="cubietech cubieboard (Cortex-A9)" ; + mc->default_cpu_type =ARM_CPU_TYPE_NAME ("cortex-a9" ); mc->init = cubieboard_init; mc->block_default_type =IF_IDE ; mc->units_per_default_bus =1 ;hw/arm/digic.c +2
-1@@ -36,7 +36,8 @@ static void digic_init(Object *obj)int i;object_initialize_child (obj,"cpu" , &s->cpu,sizeof (s->cpu), -"arm946-" TYPE_ARM_CPU , &error_abort,NULL ); +ARM_CPU_TYPE_NAME ("arm946" ), + &error_abort,NULL );for (i =0 ; i <DIGIC4_NB_TIMERS ; i++) {#define DIGIC_TIMER_NAME_MLEN 11hw/arm/fsl-imx6.c +2
-1@@ -40,7 +40,8 @@ static void fsl_imx6_init(Object *obj)for (i =0 ; i <MIN (smp_cpus,FSL_IMX6_NUM_CPUS ); i++) {snprintf (name,NAME_SIZE ,"cpu%d" , i);object_initialize_child (obj, name, &s->cpu[i],sizeof (s->cpu[i]), -"cortex-a9-" TYPE_ARM_CPU , &error_abort,NULL ); +ARM_CPU_TYPE_NAME ("cortex-a9" ), + &error_abort,NULL ); } sysbus_init_child_obj(obj, "a9mpcore", &s->a9mpcore, sizeof(s->a9mpcore),hw/arm/fsl-imx6ul.c +2
-1@@ -35,7 +35,8 @@ static void fsl_imx6ul_init(Object *obj)for (i =0 ; i <MIN (smp_cpus,FSL_IMX6UL_NUM_CPUS ); i++) {snprintf (name,NAME_SIZE ,"cpu%d" , i);object_initialize_child (obj, name, &s->cpu[i],sizeof (s->cpu[i]), -"cortex-a7-" TYPE_ARM_CPU , &error_abort,NULL ); +ARM_CPU_TYPE_NAME ("cortex-a7" ), + &error_abort,NULL ); }/* hw/arm/xlnx-zynqmp.c +4
-5@@ -193,8 +193,8 @@ static void xlnx_zynqmp_create_rpu(XlnxZynqMPState *s, const char *boot_cpu,object_initialize_child (OBJECT (&s->rpu_cluster),"rpu-cpu[*]" , &s->rpu_cpu[i],sizeof (s->rpu_cpu[i]), -"cortex-r5f-" TYPE_ARM_CPU , &error_abort, -NULL ); +ARM_CPU_TYPE_NAME ("cortex-r5f" ), + &error_abort,NULL ); name =object_get_canonical_path_component (OBJECT (&s->rpu_cpu[i]));if (strcmp (name, boot_cpu)) {@@ -233,8 +233,8 @@ static void xlnx_zynqmp_init(Object *obj)for (i =0 ; i < num_apus; i++) {object_initialize_child (OBJECT (&s->apu_cluster),"apu-cpu[*]" , &s->apu_cpu[i],sizeof (s->apu_cpu[i]), -"cortex-a53-" TYPE_ARM_CPU , &error_abort, -NULL ); +ARM_CPU_TYPE_NAME ("cortex-a53" ), + &error_abort,NULL ); } sysbus_init_child_obj(obj, "gic", &s->gic, sizeof(s->gic), --2.20.1
Re: [Qemu-devel] [PATCH 1/6] hw/arm: Use ARM_CPU_TYPE_NAME() macro when appropriate
In reply to Philippe Mathieu-DaudéQuoted 13 lines · Philippe Mathieu-Daudé
Commit
ba1ba5ccaintroduce theARM_CPU_TYPE_NAME()macro. Unify the code base by use it in all places.Signed-off-by Philippe Mathieu-Daudé <[email protected]>---
hw/arm/fsl-imx6ul.c +0
-0
hw/arm/fsl-imx6ul.c static void fsl_imx6ul_init(Object *obj) · line 42
Quoted 9 lines
hw/arm/fsl-imx6ul.c +2
-1@@ -35,7 +35,8 @@ static void fsl_imx6ul_init(Object *obj)for (i =0 ; i <MIN (smp_cpus,FSL_IMX6UL_NUM_CPUS ); i++) {snprintf (name,NAME_SIZE ,"cpu%d" , i);object_initialize_child (obj, name, &s->cpu[i],sizeof (s->cpu[i]), -"cortex-a7-" TYPE_ARM_CPU , &error_abort,NULL ); +ARM_CPU_TYPE_NAME ("cortex-a7" ), + &error_abort,NULL ); }/*
You'll find this part needs a minor fix to apply to master now we've got rid of this loop-over-CPUs in fsl-imx6ul.c.
thanks
-- PMM
PATCH 2/6 hw/arm: Use object_initialize_child for correct reference counting
In reply to Philippe Mathieu-DaudéAs explained in commit aff39be0ed97:
Both functions, object_initialize() and object_property_add_child() increase the reference counter of the new object, so one of the references has to be dropped afterwards to get the reference counting right. Otherwise the child object will not be properly cleaned up when the parent gets destroyed.
Thus let's use now object_initialize_child() instead to get the reference counting here right.
--- hw/arm/mcimx7d-sabre.c | 9 ++++----- hw/arm/mps2-tz.c | 15 +++++++-------- hw/arm/musca.c | 9 +++++---- 3 files changed, 16 insertions(+), 17 deletions(-)
hw/arm/mcimx7d-sabre.c +4 -5
@@ -29,7 +29,6 @@ static void mcimx7d_sabre_init(MachineState *machine) {static struct arm_boot_info boot_info;MCIMX7Sabre *s =g_new0 (MCIMX7Sabre,1 ); -Object *soc;int i;if (machine->ram_size >FSL_IMX7_MMDC_SIZE ) {
@@ -48,10 +47,10 @@ static void mcimx7d_sabre_init(MachineState *machine) .nb_cpus = smp_cpus, }; -object_initialize (&s->soc,sizeof (s->soc),TYPE_FSL_IMX7 ); - soc =OBJECT (&s->soc); -object_property_add_child (OBJECT (machine),"soc" , soc, &error_fatal); -object_property_set_bool (soc, true,"realized" , &error_fatal); +object_initialize_child (OBJECT (machine),"soc" , + &s->soc,sizeof (s->soc), +TYPE_FSL_IMX7 , &error_fatal,NULL ); +object_property_set_bool (OBJECT (&s->soc), true,"realized" , &error_fatal);memory_region_allocate_system_memory (&s->ram,NULL ,"mcimx7d-sabre.ram" , machine->ram_size);
hw/arm/mps2-tz.c +7 -8
@@ -427,10 +427,10 @@ static void mps2tz_common_init(MachineState *machine)/* The sec_resp_cfg output from the IoTKit must be split into multiple * lines, one for each of the PPCs we create here, plus one per MSC. */ -object_initialize (&mms->sec_resp_splitter,sizeof (mms->sec_resp_splitter), -TYPE_SPLIT_IRQ ); -object_property_add_child (OBJECT (machine),"sec-resp-splitter" , -OBJECT (&mms->sec_resp_splitter), &error_abort); +object_initialize_child (OBJECT (machine),"sec-resp-splitter" , + &mms->sec_resp_splitter, +sizeof (mms->sec_resp_splitter), +TYPE_SPLIT_IRQ , &error_abort,NULL );object_property_set_int (OBJECT (&mms->sec_resp_splitter),ARRAY_SIZE (mms->ppc) +ARRAY_SIZE (mms->msc),"num-lines" , &error_fatal);
@@ -465,10 +465,9 @@ static void mps2tz_common_init(MachineState *machine)* Tx, Rx and "combined" IRQs are sent to the NVIC separately. * Create the OR gate for this. */ -object_initialize (&mms->uart_irq_orgate,sizeof (mms->uart_irq_orgate), -TYPE_OR_IRQ ); -object_property_add_child (OBJECT (mms),"uart-irq-orgate" , -OBJECT (&mms->uart_irq_orgate), &error_abort); +object_initialize_child (OBJECT (mms),"uart-irq-orgate" , + &mms->uart_irq_orgate,sizeof (mms->uart_irq_orgate), +TYPE_OR_IRQ , &error_abort,NULL );object_property_set_int (OBJECT (&mms->uart_irq_orgate),10 ,"num-lines" , &error_fatal); object_property_set_bool(OBJECT(&mms->uart_irq_orgate), true,
hw/arm/musca.c +5 -4
@@ -424,10 +424,11 @@ static void musca_init(MachineState *machine)* The sec_resp_cfg output from the SSE-200 must be split into multiple * lines, one for each of the PPCs we create here. */ -object_initialize (&mms->sec_resp_splitter,sizeof (mms->sec_resp_splitter), -TYPE_SPLIT_IRQ ); -object_property_add_child (OBJECT (machine),"sec-resp-splitter" , -OBJECT (&mms->sec_resp_splitter), &error_fatal); +object_initialize_child (OBJECT (machine),"sec-resp-splitter" , + &mms->sec_resp_splitter, +sizeof (mms->sec_resp_splitter), +TYPE_SPLIT_IRQ , &error_fatal,NULL ); +object_property_set_int (OBJECT (&mms->sec_resp_splitter),ARRAY_SIZE (mms->ppc),"num-lines" , &error_fatal); object_property_set_bool(OBJECT(&mms->sec_resp_splitter), true,
Signature · 1 line
2.20.1
Re: [Qemu-devel] [PATCH 2/6] hw/arm: Use object_initialize_child for correct reference counting
Quoted 11 lines · Philippe Mathieu-Daudé
As explained in commit
aff39be0ed97:Both functions,
object_initialize()andobject_property_add_child()increase the reference counter of the new object, so one of the references has to be dropped afterwards to get the reference counting right. Otherwise the child object will not be properly cleaned up when the parent gets destroyed.
Thus let's use nowobject_initialize_child()instead to get the reference counting here right.
Quoted 2 lines
Signed-off-by Philippe Mathieu-Daudé <[email protected]>---
thanks
-- PMM
PATCH 3/6 hw/arm: Use sysbus_init_child_obj for correct reference counting
In reply to Philippe Mathieu-DaudéAs explained in commit aff39be0ed97:
Both functions, object_initialize() and object_property_add_child() increase the reference counter of the new object, so one of the references has to be dropped afterwards to get the reference counting right. Otherwise the child object will not be properly cleaned up when the parent gets destroyed.
Thus let's use now object_initialize_child() instead to get the reference counting here right.
--- hw/arm/exynos4_boards.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
hw/arm/exynos4_boards.c +2 -2
@@ -129,8 +129,8 @@ exynos4_boards_init_common(MachineState *machine,exynos4_boards_init_ram (s,get_system_memory (), exynos4_board_ram_size[board_type]); -object_initialize (&s->soc,sizeof (s->soc),TYPE_EXYNOS4210_SOC ); -qdev_set_parent_bus (DEVICE (&s->soc),sysbus_get_default ()); +sysbus_init_child_obj (OBJECT (machine),"soc" , + &s->soc,sizeof (s->soc),TYPE_EXYNOS4210_SOC );object_property_set_bool (OBJECT (&s->soc), true,"realized" , &error_fatal);
Signature · 1 line
2.20.1
Re: [Qemu-arm] [PATCH 3/6] hw/arm: Use sysbus_init_child_obj for correct reference counting
Quoted 21 lines · Philippe Mathieu-Daudé
As explained in commit
aff39be0ed97:Both functions,
object_initialize()andobject_property_add_child()increase the reference counter of the new object, so one of the references has to be dropped afterwards to get the reference counting right. Otherwise the child object will not be properly cleaned up when the parent gets destroyed.
Thus let's use nowobject_initialize_child()instead to get the reference counting here right.Signed-off-by Philippe Mathieu-Daudé <[email protected]>--- hw/arm/exynos4_boards.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)hw/arm/exynos4_boards.c +0
-0
hw/arm/exynos4_boards.c exynos4_boards_init_common(MachineState *machine, · line 135
Quoted 9 lines
hw/arm/exynos4_boards.c +2
-2@@ -129,8 +129,8 @@ exynos4_boards_init_common(MachineState *machine,exynos4_boards_init_ram (s,get_system_memory (), exynos4_board_ram_size[board_type]); -object_initialize (&s->soc,sizeof (s->soc),TYPE_EXYNOS4210_SOC ); -qdev_set_parent_bus (DEVICE (&s->soc),sysbus_get_default ()); +sysbus_init_child_obj (OBJECT (machine),"soc" , + &s->soc,sizeof (s->soc),TYPE_EXYNOS4210_SOC );object_property_set_bool (OBJECT (&s->soc), true,"realized" , &error_fatal);
I suspect the code change here is correct but it doesn't seem to match the commit message -- the old code is not calling object_property_add_child() at all, and the new code does not call object_initialize_child()...
thanks
-- PMM
Re: [Qemu-arm] [PATCH 3/6] hw/arm: Use sysbus_init_child_obj for correct reference counting
Quoted 25 lines · Peter Maydell
Quoted 24 lines · Philippe Mathieu-Daudé
As explained in commit
aff39be0ed97:Both functions,
object_initialize()andobject_property_add_child()increase the reference counter of the new object, so one of the references has to be dropped afterwards to get the reference counting right. Otherwise the child object will not be properly cleaned up when the parent gets destroyed.
Thus let's use nowobject_initialize_child()instead to get the reference counting here right.Signed-off-by Philippe Mathieu-Daudé <[email protected]>--- hw/arm/exynos4_boards.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)hw/arm/exynos4_boards.c +0
-0@@ -129,8 +129,8 @@ exynos4_boards_init_common(MachineState *machine,exynos4_boards_init_ram (s,get_system_memory (), exynos4_board_ram_size[board_type]);
Quoted 11 lines
hw/arm/exynos4_boards.c exynos4_boards_init_common(MachineState *machine, · line 135
Quoted 6 lines
hw/arm/exynos4_boards.c +2
-2-object_initialize (&s->soc,sizeof (s->soc),TYPE_EXYNOS4210_SOC ); -qdev_set_parent_bus (DEVICE (&s->soc),sysbus_get_default ()); +sysbus_init_child_obj (OBJECT (machine),"soc" , + &s->soc,sizeof (s->soc),TYPE_EXYNOS4210_SOC );object_property_set_bool (OBJECT (&s->soc), true,"realized" , &error_fatal);I suspect the code change here is correct but it doesn't seem to match the commit message -- the old code is not calling
object_property_add_child()at all, and the new code does not callobject_initialize_child()...
OK, will improve, thanks!
PATCH 4/6 hw/arm/fsl-imx: Add the cpu as child of the SoC object
In reply to Philippe Mathieu-Daudé--- hw/arm/fsl-imx25.c | 4 +++- hw/arm/fsl-imx31.c | 4 +++- 2 files changed, 6 insertions(+), 2 deletions(-)
hw/arm/fsl-imx25.c +3 -1
@@ -36,7 +36,9 @@ static void fsl_imx25_init(Object *obj)FslIMX25State *s =FSL_IMX25 (obj);int i; -object_initialize (&s->cpu,sizeof (s->cpu),"arm926-" TYPE_ARM_CPU ); +object_initialize_child (obj,"cpu" , &s->cpu,sizeof (s->cpu), +ARM_CPU_TYPE_NAME ("arm926" ), + &error_abort,NULL );sysbus_init_child_obj (obj,"avic" , &s->avic,sizeof (s->avic),TYPE_IMX_AVIC );
hw/arm/fsl-imx31.c +3 -1
@@ -33,7 +33,9 @@ static void fsl_imx31_init(Object *obj)FslIMX31State *s =FSL_IMX31 (obj);int i; -object_initialize (&s->cpu,sizeof (s->cpu),"arm1136-" TYPE_ARM_CPU ); +object_initialize_child (obj,"cpu" , &s->cpu,sizeof (s->cpu), +ARM_CPU_TYPE_NAME ("arm1136" ), + &error_abort,NULL );sysbus_init_child_obj (obj,"avic" , &s->avic,sizeof (s->avic),TYPE_IMX_AVIC );
Signature · 1 line
2.20.1
Re: [Qemu-arm] [PATCH 4/6] hw/arm/fsl-imx: Add the cpu as child of the SoC object
Quoted 38 lines · Philippe Mathieu-Daudé
Signed-off-by Philippe Mathieu-Daudé <[email protected]>--- hw/arm/fsl-imx25.c | 4 +++- hw/arm/fsl-imx31.c | 4 +++- 2 files changed, 6 insertions(+), 2 deletions(-)hw/arm/fsl-imx25.c +3
-1@@ -36,7 +36,9 @@ static void fsl_imx25_init(Object *obj)FslIMX25State *s =FSL_IMX25 (obj);int i; -object_initialize (&s->cpu,sizeof (s->cpu),"arm926-" TYPE_ARM_CPU ); +object_initialize_child (obj,"cpu" , &s->cpu,sizeof (s->cpu), +ARM_CPU_TYPE_NAME ("arm926" ), + &error_abort,NULL );sysbus_init_child_obj (obj,"avic" , &s->avic,sizeof (s->avic),TYPE_IMX_AVIC );hw/arm/fsl-imx31.c +3
-2@@ -33,7 +33,9 @@ static void fsl_imx31_init(Object *obj)FslIMX31State *s =FSL_IMX31 (obj);int i; -object_initialize (&s->cpu,sizeof (s->cpu),"arm1136-" TYPE_ARM_CPU ); +object_initialize_child (obj,"cpu" , &s->cpu,sizeof (s->cpu), +ARM_CPU_TYPE_NAME ("arm1136" ), + &error_abort,NULL );sysbus_init_child_obj (obj,"avic" , &s->avic,sizeof (s->avic),TYPE_IMX_AVIC ); --
Quoted 1 line
2.20.1
Really the ARM_CPU_TYPE_NAME part of this change should be in patch 1 I think...
Could you expand the commit message a little to explain why we want to make the CPU a child of the SoC object?
thanks
-- PMM
PATCH 5/6 hw/dma/xilinx_axi: Use object_initialize_child for correct ref. counting
In reply to Philippe Mathieu-DaudéAs explained in commit aff39be0ed97:
Both functions, object_initialize() and object_property_add_child() increase the reference counter of the new object, so one of the references has to be dropped afterwards to get the reference counting right. Otherwise the child object will not be properly cleaned up when the parent gets destroyed.
Thus let's use now object_initialize_child() instead to get the reference counting here right.
--- hw/dma/xilinx_axidma.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-)
hw/dma/xilinx_axidma.c +8 -8
@@ -563,14 +563,14 @@ static void xilinx_axidma_init(Object *obj)XilinxAXIDMA *s =XILINX_AXI_DMA (obj);SysBusDevice *sbd =SYS_BUS_DEVICE (obj); -object_initialize (&s->rx_data_dev,sizeof (s->rx_data_dev), -TYPE_XILINX_AXI_DMA_DATA_STREAM ); -object_initialize (&s->rx_control_dev,sizeof (s->rx_control_dev), -TYPE_XILINX_AXI_DMA_CONTROL_STREAM ); -object_property_add_child (OBJECT (s),"axistream-connected-target" , - (Object *)&s->rx_data_dev, &error_abort); -object_property_add_child (OBJECT (s),"axistream-control-connected-target" , - (Object *)&s->rx_control_dev, &error_abort); +object_initialize_child (OBJECT (s),"axistream-connected-target" , + &s->rx_data_dev,sizeof (s->rx_data_dev), +TYPE_XILINX_AXI_ENET_DATA_STREAM , &error_abort, +NULL ); +object_initialize_child (OBJECT (s),"axistream-control-connected-target" , + &s->rx_control_dev,sizeof (s->rx_control_dev), +TYPE_XILINX_AXI_ENET_CONTROL_STREAM , &error_abort, +NULL );sysbus_init_irq (sbd, &s->streams[0 ].irq);sysbus_init_irq (sbd, &s->streams[1 ].irq);
Signature · 1 line
2.20.1
Re: [Qemu-devel] [PATCH 5/6] hw/dma/xilinx_axi: Use object_initialize_child for correct ref. counting
Quoted 11 lines · Philippe Mathieu-Daudé
As explained in commit
aff39be0ed97:Both functions,
object_initialize()andobject_property_add_child()increase the reference counter of the new object, so one of the references has to be dropped afterwards to get the reference counting right. Otherwise the child object will not be properly cleaned up when the parent gets destroyed.
Thus let's use nowobject_initialize_child()instead to get the reference counting here right.
Quoted 1 line
Signed-off-by Philippe Mathieu-Daudé <[email protected]>
Alistair
Quoted 35 lines
--- hw/dma/xilinx_axidma.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-)hw/dma/xilinx_axidma.c +8
-9@@ -563,14 +563,14 @@ static void xilinx_axidma_init(Object *obj)XilinxAXIDMA *s =XILINX_AXI_DMA (obj);SysBusDevice *sbd =SYS_BUS_DEVICE (obj); -object_initialize (&s->rx_data_dev,sizeof (s->rx_data_dev), -TYPE_XILINX_AXI_DMA_DATA_STREAM ); -object_initialize (&s->rx_control_dev,sizeof (s->rx_control_dev), -TYPE_XILINX_AXI_DMA_CONTROL_STREAM ); -object_property_add_child (OBJECT (s),"axistream-connected-target" , - (Object *)&s->rx_data_dev, &error_abort); -object_property_add_child (OBJECT (s),"axistream-control-connected-target" , - (Object *)&s->rx_control_dev, &error_abort); +object_initialize_child (OBJECT (s),"axistream-connected-target" , + &s->rx_data_dev,sizeof (s->rx_data_dev), +TYPE_XILINX_AXI_ENET_DATA_STREAM , &error_abort, +NULL ); +object_initialize_child (OBJECT (s),"axistream-control-connected-target" , + &s->rx_control_dev,sizeof (s->rx_control_dev), +TYPE_XILINX_AXI_ENET_CONTROL_STREAM , &error_abort, +NULL );sysbus_init_irq (sbd, &s->streams[0 ].irq);sysbus_init_irq (sbd, &s->streams[1 ].irq); --2.20.1
Re: [Qemu-arm] [Qemu-devel] [PATCH 5/6] hw/dma/xilinx_axi: Use object_initialize_child for correct ref. counting
Quoted 47 lines · Alistair Francis
Quoted 11 lines · Philippe Mathieu-Daudé
As explained in commit
aff39be0ed97:Both functions,
object_initialize()andobject_property_add_child()increase the reference counter of the new object, so one of the references has to be dropped afterwards to get the reference counting right. Otherwise the child object will not be properly cleaned up when the parent gets destroyed.
Thus let's use nowobject_initialize_child()instead to get the reference counting here right.Quoted 1 line
Signed-off-by Philippe Mathieu-Daudé <[email protected]>Reviewed-by Alistair Francis <[email protected]>Alistair
Quoted 29 lines
--- hw/dma/xilinx_axidma.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-)hw/dma/xilinx_axidma.c +8
-8@@ -563,14 +563,14 @@ static void xilinx_axidma_init(Object *obj)XilinxAXIDMA *s =XILINX_AXI_DMA (obj);SysBusDevice *sbd =SYS_BUS_DEVICE (obj); -object_initialize (&s->rx_data_dev,sizeof (s->rx_data_dev), -TYPE_XILINX_AXI_DMA_DATA_STREAM ); -object_initialize (&s->rx_control_dev,sizeof (s->rx_control_dev), -TYPE_XILINX_AXI_DMA_CONTROL_STREAM ); -object_property_add_child (OBJECT (s),"axistream-connected-target" , - (Object *)&s->rx_data_dev, &error_abort); -object_property_add_child (OBJECT (s),"axistream-control-connected-target" , - (Object *)&s->rx_control_dev, &error_abort); +object_initialize_child (OBJECT (s),"axistream-connected-target" , + &s->rx_data_dev,sizeof (s->rx_data_dev), +TYPE_XILINX_AXI_ENET_DATA_STREAM , &error_abort, +NULL ); +object_initialize_child (OBJECT (s),"axistream-control-connected-target" , + &s->rx_control_dev,sizeof (s->rx_control_dev), +TYPE_XILINX_AXI_ENET_CONTROL_STREAM , &error_abort, +NULL );
Quoted 4 lines
Quoted 4 lines
hw/dma/xilinx_axidma.c +0
-1sysbus_init_irq (sbd, &s->streams[0 ].irq);sysbus_init_irq (sbd, &s->streams[1 ].irq); --2.20.1
As GCC suggested, this patch requires this snippet squashed:
-- >8 --
hw/dma/xilinx_axidma.c +1 -1
@@ -565,11 +565,11 @@ static void xilinx_axidma_init(Object *obj)object_initialize_child (OBJECT (s),"axistream-connected-target" , &s->rx_data_dev,sizeof (s->rx_data_dev), -TYPE_XILINX_AXI_ENET_DATA_STREAM , &error_abort, +TYPE_XILINX_AXI_DMA_DATA_STREAM , &error_abort,NULL ); object_initialize_child(OBJECT(s),
"axistream-control-connected-target",
&s->rx_control_dev, sizeof(s->rx_control_dev),
- TYPE_XILINX_AXI_ENET_CONTROL_STREAM,&error_abort,
+ TYPE_XILINX_AXI_DMA_CONTROL_STREAM,&error_abort,
NULL);
sysbus_init_irq(sbd, &s->streams[0].irq);---
Since it was commented in reply to the cover, I assume your R-b stands with the snippet applied.
Thanks for reviewing,
Phil.
Re: [Qemu-devel] [PATCH 5/6] hw/dma/xilinx_axi: Use object_initialize_child for correct ref. counting
Quoted 75 lines · Philippe Mathieu-Daudé
Quoted 47 lines · Alistair Francis
Quoted 11 lines · Philippe Mathieu-Daudé
As explained in commit
aff39be0ed97:Both functions,
object_initialize()andobject_property_add_child()increase the reference counter of the new object, so one of the references has to be dropped afterwards to get the reference counting right. Otherwise the child object will not be properly cleaned up when the parent gets destroyed.
Thus let's use nowobject_initialize_child()instead to get the reference counting here right.Quoted 1 line
Signed-off-by Philippe Mathieu-Daudé <[email protected]>Reviewed-by Alistair Francis <[email protected]>Alistair
Quoted 29 lines
--- hw/dma/xilinx_axidma.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-)hw/dma/xilinx_axidma.c +8
-8@@ -563,14 +563,14 @@ static void xilinx_axidma_init(Object *obj)XilinxAXIDMA *s =XILINX_AXI_DMA (obj);SysBusDevice *sbd =SYS_BUS_DEVICE (obj); -object_initialize (&s->rx_data_dev,sizeof (s->rx_data_dev), -TYPE_XILINX_AXI_DMA_DATA_STREAM ); -object_initialize (&s->rx_control_dev,sizeof (s->rx_control_dev), -TYPE_XILINX_AXI_DMA_CONTROL_STREAM ); -object_property_add_child (OBJECT (s),"axistream-connected-target" , - (Object *)&s->rx_data_dev, &error_abort); -object_property_add_child (OBJECT (s),"axistream-control-connected-target" , - (Object *)&s->rx_control_dev, &error_abort); +object_initialize_child (OBJECT (s),"axistream-connected-target" , + &s->rx_data_dev,sizeof (s->rx_data_dev), +TYPE_XILINX_AXI_ENET_DATA_STREAM , &error_abort, +NULL ); +object_initialize_child (OBJECT (s),"axistream-control-connected-target" , + &s->rx_control_dev,sizeof (s->rx_control_dev), +TYPE_XILINX_AXI_ENET_CONTROL_STREAM , &error_abort, +NULL );Quoted 4 lines
Quoted 4 lines
hw/dma/xilinx_axidma.c +0
-1sysbus_init_irq (sbd, &s->streams[0 ].irq);sysbus_init_irq (sbd, &s->streams[1 ].irq); --2.20.1
As GCC suggested, this patch requires this snippet squashed:
-- >8 --
hw/dma/xilinx_axidma.c +2
-2@@ -565,11 +565,11 @@ static void xilinx_axidma_init(Object *obj)object_initialize_child (OBJECT (s),"axistream-connected-target" , &s->rx_data_dev,sizeof (s->rx_data_dev), -TYPE_XILINX_AXI_ENET_DATA_STREAM , &error_abort, +TYPE_XILINX_AXI_DMA_DATA_STREAM , &error_abort,NULL ); object_initialize_child(OBJECT(s), "axistream-control-connected-target", &s->rx_control_dev, sizeof(s->rx_control_dev), - TYPE_XILINX_AXI_ENET_CONTROL_STREAM, &error_abort, + TYPE_XILINX_AXI_DMA_CONTROL_STREAM, &error_abort,NULL); sysbus_init_irq(sbd, &s->streams[0].irq);
Quoted 1 line
---
Ah I missed that
Quoted 3 lines
Since it was commented in reply to the cover, I assume your R-b stands with the snippet applied.
Yeah it does.
Alistair
Quoted 4 lines
Thanks for reviewing,
Phil.
Buh I forgot to Cc Eduardo.
Quoted 24 lines · Philippe Mathieu-Daudé
First we use
ARM_CPU_TYPE_NAME()when we should.Then is follow up of
https://lists.gnu.org/archive/html/qemu-devel/2019-05/msg01492.htmlThis series looks at Eduardo suggestions from [1] and Thomas commit
aff39be0ed97to replace variousobject_initialize+qdev_set_parent_buscalls bysysbus_init_child_obj().Finally, some devices are declared orphean while they have a parent, let them be together again.
[1] https://patchwork.ozlabs.org/patch/943333/#1953608
Philippe Mathieu-Daudé (6):
hw/arm: UseARM_CPU_TYPE_NAME()macro when appropriate hw/arm: Useobject_initialize_childfor correct reference counting hw/arm: Usesysbus_init_child_objfor correct reference counting hw/arm/fsl-imx: Add the cpu as child of the SoC object hw/dma/xilinx_axi: Useobject_initialize_childfor correct ref.
counting
hw/net/xilinx_axi: Useobject_initialize_childfor correct ref.
counting
PATCH 6/6 hw/net/xilinx_axi: Use object_initialize_child for correct ref. counting
In reply to Philippe Mathieu-DaudéAs explained in commit aff39be0ed97:
Both functions, object_initialize() and object_property_add_child() increase the reference counter of the new object, so one of the references has to be dropped afterwards to get the reference counting right. Otherwise the child object will not be properly cleaned up when the parent gets destroyed.
Thus let's use now object_initialize_child() instead to get the reference counting here right.
--- hw/net/xilinx_axienet.c | 17 ++++++++--------- 1 file changed, 8 insertions(+), 9 deletions(-)
hw/net/xilinx_axienet.c +8 -9
@@ -990,15 +990,14 @@ static void xilinx_enet_init(Object *obj)XilinxAXIEnet *s =XILINX_AXI_ENET (obj);SysBusDevice *sbd =SYS_BUS_DEVICE (obj); -object_initialize (&s->rx_data_dev,sizeof (s->rx_data_dev), -TYPE_XILINX_AXI_ENET_DATA_STREAM ); -object_initialize (&s->rx_control_dev,sizeof (s->rx_control_dev), -TYPE_XILINX_AXI_ENET_CONTROL_STREAM ); -object_property_add_child (OBJECT (s),"axistream-connected-target" , - (Object *)&s->rx_data_dev, &error_abort); -object_property_add_child (OBJECT (s),"axistream-control-connected-target" , - (Object *)&s->rx_control_dev, &error_abort); - +object_initialize_child (OBJECT (s),"axistream-connected-target" , + &s->rx_data_dev,sizeof (s->rx_data_dev), +TYPE_XILINX_AXI_ENET_DATA_STREAM , &error_abort, +NULL ); +object_initialize_child (OBJECT (s),"axistream-control-connected-target" , + &s->rx_control_dev,sizeof (s->rx_control_dev), +TYPE_XILINX_AXI_ENET_CONTROL_STREAM , &error_abort, +NULL );sysbus_init_irq (sbd, &s->irq);memory_region_init_io (&s->iomem,OBJECT (s), &enet_ops, s,"enet" ,0x40000 );
Signature · 1 line
2.20.1
Re: [Qemu-devel] [PATCH 6/6] hw/net/xilinx_axi: Use object_initialize_child for correct ref. counting
Quoted 11 lines · Philippe Mathieu-Daudé
As explained in commit
aff39be0ed97:Both functions,
object_initialize()andobject_property_add_child()increase the reference counter of the new object, so one of the references has to be dropped afterwards to get the reference counting right. Otherwise the child object will not be properly cleaned up when the parent gets destroyed.
Thus let's use nowobject_initialize_child()instead to get the reference counting here right.
Quoted 1 line
Signed-off-by Philippe Mathieu-Daudé <[email protected]>
Alistair
Quoted 36 lines
--- hw/net/xilinx_axienet.c | 17 ++++++++--------- 1 file changed, 8 insertions(+), 9 deletions(-)hw/net/xilinx_axienet.c +8
-10@@ -990,15 +990,14 @@ static void xilinx_enet_init(Object *obj)XilinxAXIEnet *s =XILINX_AXI_ENET (obj);SysBusDevice *sbd =SYS_BUS_DEVICE (obj); -object_initialize (&s->rx_data_dev,sizeof (s->rx_data_dev), -TYPE_XILINX_AXI_ENET_DATA_STREAM ); -object_initialize (&s->rx_control_dev,sizeof (s->rx_control_dev), -TYPE_XILINX_AXI_ENET_CONTROL_STREAM ); -object_property_add_child (OBJECT (s),"axistream-connected-target" , - (Object *)&s->rx_data_dev, &error_abort); -object_property_add_child (OBJECT (s),"axistream-control-connected-target" , - (Object *)&s->rx_control_dev, &error_abort); - +object_initialize_child (OBJECT (s),"axistream-connected-target" , + &s->rx_data_dev,sizeof (s->rx_data_dev), +TYPE_XILINX_AXI_ENET_DATA_STREAM , &error_abort, +NULL ); +object_initialize_child (OBJECT (s),"axistream-control-connected-target" , + &s->rx_control_dev,sizeof (s->rx_control_dev), +TYPE_XILINX_AXI_ENET_CONTROL_STREAM , &error_abort, +NULL );sysbus_init_irq (sbd, &s->irq);memory_region_init_io (&s->iomem,OBJECT (s), &enet_ops, s,"enet" ,0x40000 ); --2.20.1
Patchew URL: https://patchew.org/QEMU/[email protected]/
Patchew URL: https://patchew.org/QEMU/[email protected]/
Hi,
This series failed build test on s390x host. Please find the details below.
=== TEST SCRIPT BEGIN ===
#!/bin/bash
# Testing script will be invoked under the git checkout with # HEAD pointing to a commit that has the patches applied on top of "base"
# branch set -e
echo
echo "=== ENV ==="
env
echo
echo "=== PACKAGES ==="
rpm -qa
echo
echo "=== UNAME ==="
uname -a
CC=$HOME/bin/cc
INSTALL=$PWD/install
BUILD=$PWD/build
mkdir -p $BUILD $INSTALL
SRC=$PWD
cd $BUILD
$SRC/configure --cc=$CC --prefix=$INSTALL
make -j4
# XXX: we need reliable clean up
# make check -j4 V=1 make install === TEST SCRIPT END ===
CC hw/dma/i8257.o
CC hw/dma/xilinx_axidma.o/var/tmp/patchew-tester-tmp-6umc5eop/src/hw/dma/xilinx_axidma.c: In function ‘xilinx_axidma_init’:/var/tmp/patchew-tester-tmp-6umc5eop/src/hw/dma/xilinx_axidma.c:568:29: error: ‘TYPE_XILINX_AXI_ENET_DATA_STREAM’ undeclared (first use in this function); did you mean ‘TYPE_XILINX_AXI_DMA_DATA_STREAM’?
568 | TYPE_XILINX_AXI_ENET_DATA_STREAM, &error_abort,
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
| TYPE_XILINX_AXI_DMA_DATA_STREAM
/var/tmp/patchew-tester-tmp-6umc5eop/src/hw/dma/xilinx_axidma.c:568:29: note: each undeclared identifier is reported only once for each function it appears in
/var/tmp/patchew-tester-tmp-6umc5eop/src/hw/dma/xilinx_axidma.c:572:29: error: ‘TYPE_XILINX_AXI_ENET_CONTROL_STREAM’ undeclared (first use in this function); did you mean ‘TYPE_XILINX_AXI_DMA_CONTROL_STREAM’?
572 | TYPE_XILINX_AXI_ENET_CONTROL_STREAM, &error_abort,
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
| TYPE_XILINX_AXI_DMA_CONTROL_STREAMThe full log is available at
http://patchew.org/logs/[email protected]/testing.s390x/?type=message.
---
Email generated automatically by Patchew [https://patchew.org/].
Please send your feedback to [email protected]
Quoted 10 lines · On 7/1/19 7:56 PM, [email protected] wrote:
Patchew URL: https://patchew.org/QEMU/[email protected]/
Hi,
This series failed build test on s390x host. Please find the details below.
CC
hw/dma/i8257.o
CChw/dma/xilinx_axidma.o/var/tmp/patchew-tester-tmp-6umc5eop/src/hw/dma/xilinx_axidma.c: In function ‘xilinx_axidma_init’:/var/tmp/patchew-tester-tmp-6umc5eop/src/hw/dma/xilinx_axidma.c:568:29: error: ‘TYPE_XILINX_AXI_ENET_DATA_STREAM’ undeclared (first use in this function); did you mean ‘TYPE_XILINX_AXI_DMA_DATA_STREAM’?
Quoted 8 lines
568 | TYPE_XILINX_AXI_ENET_DATA_STREAM, &error_abort,
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
| TYPE_XILINX_AXI_DMA_DATA_STREAM
/var/tmp/patchew-tester-tmp-6umc5eop/src/hw/dma/xilinx_axidma.c:568:29: note: each undeclared identifier is reported only once for each function it appears in
/var/tmp/patchew-tester-tmp-6umc5eop/src/hw/dma/xilinx_axidma.c:572:29: error: ‘TYPE_XILINX_AXI_ENET_CONTROL_STREAM’ undeclared (first use in this function); did you mean ‘TYPE_XILINX_AXI_DMA_CONTROL_STREAM’?
572 | TYPE_XILINX_AXI_ENET_CONTROL_STREAM, &error_abort,
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
| TYPE_XILINX_AXI_DMA_CONTROL_STREAMIt would be nice if GCC directly fix the patch instead of embarrassing the author for obvious copy/paste mistakes...
Quoted 21 lines · Philippe Mathieu-Daudé
Quoted 10 lines · On 7/1/19 7:56 PM, [email protected] wrote:
Patchew URL: https://patchew.org/QEMU/[email protected]/
Hi,
This series failed build test on s390x host. Please find the details below.
CC
hw/dma/i8257.o
CChw/dma/xilinx_axidma.o/var/tmp/patchew-tester-tmp-6umc5eop/src/hw/dma/xilinx_axidma.c: In function ‘xilinx_axidma_init’:/var/tmp/patchew-tester-tmp-6umc5eop/src/hw/dma/xilinx_axidma.c:568:29: error: ‘TYPE_XILINX_AXI_ENET_DATA_STREAM’ undeclared (first use in this function); did you mean ‘TYPE_XILINX_AXI_DMA_DATA_STREAM’?Quoted 8 lines
568 | TYPE_XILINX_AXI_ENET_DATA_STREAM, &error_abort, | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | TYPE_XILINX_AXI_DMA_DATA_STREAM /var/tmp/patchew-tester-tmp-6umc5eop/src/hw/dma/xilinx_axidma.c:568:29: note: each undeclared identifier is reported only once for each function it appears in /var/tmp/patchew-tester-tmp-6umc5eop/src/hw/dma/xilinx_axidma.c:572:29: error: ‘TYPE_XILINX_AXI_ENET_CONTROL_STREAM’ undeclared (first use in this function); did you mean ‘TYPE_XILINX_AXI_DMA_CONTROL_STREAM’? 572 | TYPE_XILINX_AXI_ENET_CONTROL_STREAM, &error_abort, | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | TYPE_XILINX_AXI_DMA_CONTROL_STREAM
Quoted 2 lines
It would be nice if GCC directly fix the patch instead of embarrassing the author for obvious copy/paste mistakes...
Sufficiently new gcc supports -fdiagnostics-generate-patch, and you could probably script up a wrapper to automatically apply the patches ;-)
thanks
-- PMM