1
Arm patch queue -- these are all bug fix patches but we might
1
target-arm queue. The big thing here is the landing of the 3-phase
2
as well put them in to rc0...
2
reset patches...
3
3
4
thanks
5
-- PMM
4
-- PMM
6
5
7
The following changes since commit 2c8cfc0b52b5a4d123c26c0b5fdf941be24805be:
6
The following changes since commit 204aa60b37c23a89e690d418f49787d274303ca7:
8
7
9
Merge remote-tracking branch 'remotes/kevin/tags/for-upstream' into staging (2018-03-19 11:44:26 +0000)
8
Merge remote-tracking branch 'remotes/amarkovic/tags/mips-queue-jan-29-2020' into staging (2020-01-30 14:18:45 +0000)
10
9
11
are available in the Git repository at:
10
are available in the Git repository at:
12
11
13
git://git.linaro.org/people/pmaydell/qemu-arm.git tags/pull-target-arm-20180319
12
https://git.linaro.org/people/pmaydell/qemu-arm.git tags/pull-target-arm-20200130
14
13
15
for you to fetch changes up to ff72cb6b46b95bb530787add5277c211af3d31c6:
14
for you to fetch changes up to dea101a1ae9968c9fec6ab0291489dad7c49f36f:
16
15
17
hw/arm/raspi: Provide spin-loop code for AArch64 CPUs (2018-03-19 18:23:24 +0000)
16
target/arm/cpu: Add the kvm-no-adjvtime CPU property (2020-01-30 16:02:06 +0000)
18
17
19
----------------------------------------------------------------
18
----------------------------------------------------------------
20
target-arm queue:
19
target-arm queue:
21
* fsl-imx6: Fix incorrect Ethernet interrupt defines
20
* hw/core/or-irq: Fix incorrect assert forbidding num-lines == MAX_OR_LINES
22
* dump: Update correct kdump phys_base field for AArch64
21
* target/arm/arm-semi: Don't let the guest close stdin/stdout/stderr
23
* char: i.MX: Add support for "TX complete" interrupt
22
* aspeed: some minor bugfixes
24
* bcm2836/raspi: Fix various bugs resulting in panics trying
23
* aspeed: add eMMC controller model for AST2600 SoC
25
to boot a Debian Linux kernel on raspi3
24
* hw/arm/raspi: Remove obsolete use of -smp to set the soc 'enabled-cpus'
25
* New 3-phase reset API for device models
26
* hw/intc/arm_gicv3_kvm: Stop wrongly programming GICR_PENDBASER.PTZ bit
27
* Arm KVM: stop/restart the guest counter when the VM is stopped and started
26
28
27
----------------------------------------------------------------
29
----------------------------------------------------------------
28
Andrey Smirnov (2):
30
Andrew Jeffery (2):
29
char: i.MX: Simplify imx_update()
31
hw/sd: Configure number of slots exposed by the ASPEED SDHCI model
30
char: i.MX: Add support for "TX complete" interrupt
32
hw/arm: ast2600: Wire up the eMMC controller
31
33
32
Guenter Roeck (1):
34
Andrew Jones (6):
33
fsl-imx6: Swap Ethernet interrupt defines
35
target/arm/kvm: trivial: Clean up header documentation
36
hw/arm/virt: Add missing 5.0 options call to 4.2 options
37
target/arm/kvm64: kvm64 cpus have timer registers
38
tests/arm-cpu-features: Check feature default values
39
target/arm/kvm: Implement virtual time adjustment
40
target/arm/cpu: Add the kvm-no-adjvtime CPU property
34
41
35
Peter Maydell (9):
42
Cédric Le Goater (2):
36
hw/arm/raspi: Don't do board-setup or secure-boot for raspi3
43
ftgmac100: check RX and TX buffer alignment
37
hw/arm/boot: assert that secure_boot and secure_board_setup are false for AArch64
44
hw/arm/aspeed: add a 'execute-in-place' property to boot directly from CE0
38
hw/arm/boot: If booting a kernel in EL2, set SCR_EL3.HCE
39
hw/arm/bcm2386: Fix parent type of bcm2386
40
hw/arm/bcm2836: Rename bcm2836 type/struct to bcm283x
41
hw/arm/bcm2836: Create proper bcm2837 device
42
hw/arm/bcm2836: Use correct affinity values for BCM2837
43
hw/arm/bcm2836: Hardcode correct CPU type
44
hw/arm/raspi: Provide spin-loop code for AArch64 CPUs
45
45
46
Wei Huang (1):
46
Damien Hedde (11):
47
dump: Update correct kdump phys_base field for AArch64
47
add device_legacy_reset function to prepare for reset api change
48
hw/core/qdev: add trace events to help with resettable transition
49
hw/core: create Resettable QOM interface
50
hw/core: add Resettable support to BusClass and DeviceClass
51
hw/core/resettable: add support for changing parent
52
hw/core/qdev: handle parent bus change regarding resettable
53
hw/core/qdev: update hotplug reset regarding resettable
54
hw/core: deprecate old reset functions and introduce new ones
55
docs/devel/reset.rst: add doc about Resettable interface
56
vl: replace deprecated qbus_reset_all registration
57
hw/s390x/ipl: replace deprecated qdev_reset_all registration
48
58
49
include/hw/arm/bcm2836.h | 31 +++++++++++++---
59
Joel Stanley (1):
50
include/hw/arm/fsl-imx6.h | 4 +-
60
misc/pca9552: Add qom set and get
51
include/hw/char/imx_serial.h | 3 ++
52
dump.c | 14 +++++--
53
hw/arm/bcm2836.c | 87 +++++++++++++++++++++++++++++++-------------
54
hw/arm/boot.c | 12 ++++++
55
hw/arm/raspi.c | 77 +++++++++++++++++++++++++++++++--------
56
hw/char/imx_serial.c | 44 ++++++++++++++++------
57
hw/net/imx_fec.c | 28 +++++++++++++-
58
9 files changed, 237 insertions(+), 63 deletions(-)
59
61
62
Peter Maydell (2):
63
hw/core/or-irq: Fix incorrect assert forbidding num-lines == MAX_OR_LINES
64
target/arm/arm-semi: Don't let the guest close stdin/stdout/stderr
65
66
Philippe Mathieu-Daudé (1):
67
hw/arm/raspi: Remove obsolete use of -smp to set the soc 'enabled-cpus'
68
69
Zenghui Yu (1):
70
hw/intc/arm_gicv3_kvm: Stop wrongly programming GICR_PENDBASER.PTZ bit
71
72
hw/core/Makefile.objs | 1 +
73
tests/Makefile.include | 1 +
74
include/hw/arm/aspeed.h | 2 +
75
include/hw/arm/aspeed_soc.h | 2 +
76
include/hw/arm/virt.h | 1 +
77
include/hw/qdev-core.h | 58 +++++++-
78
include/hw/resettable.h | 247 +++++++++++++++++++++++++++++++++
79
include/hw/sd/aspeed_sdhci.h | 1 +
80
target/arm/cpu.h | 7 +
81
target/arm/kvm_arm.h | 95 ++++++++++---
82
hw/arm/aspeed.c | 72 ++++++++--
83
hw/arm/aspeed_ast2600.c | 31 ++++-
84
hw/arm/aspeed_soc.c | 2 +
85
hw/arm/raspi.c | 2 -
86
hw/arm/virt.c | 9 ++
87
hw/audio/intel-hda.c | 2 +-
88
hw/core/bus.c | 102 ++++++++++++++
89
hw/core/or-irq.c | 2 +-
90
hw/core/qdev.c | 160 ++++++++++++++++++++--
91
hw/core/resettable.c | 301 +++++++++++++++++++++++++++++++++++++++++
92
hw/hyperv/hyperv.c | 2 +-
93
hw/i386/microvm.c | 2 +-
94
hw/i386/pc.c | 2 +-
95
hw/ide/microdrive.c | 8 +-
96
hw/intc/arm_gicv3_kvm.c | 11 +-
97
hw/intc/spapr_xive.c | 2 +-
98
hw/misc/pca9552.c | 90 ++++++++++++
99
hw/net/ftgmac100.c | 13 ++
100
hw/ppc/pnv_psi.c | 4 +-
101
hw/ppc/spapr_pci.c | 2 +-
102
hw/ppc/spapr_vio.c | 2 +-
103
hw/s390x/ipl.c | 10 +-
104
hw/s390x/s390-pci-inst.c | 2 +-
105
hw/scsi/vmw_pvscsi.c | 2 +-
106
hw/sd/aspeed_sdhci.c | 11 +-
107
hw/sd/omap_mmc.c | 2 +-
108
hw/sd/pl181.c | 2 +-
109
target/arm/arm-semi.c | 9 ++
110
target/arm/cpu.c | 2 +
111
target/arm/cpu64.c | 1 +
112
target/arm/kvm.c | 120 ++++++++++++++++
113
target/arm/kvm32.c | 3 +
114
target/arm/kvm64.c | 4 +
115
target/arm/machine.c | 7 +
116
target/arm/monitor.c | 1 +
117
tests/qtest/arm-cpu-features.c | 41 ++++--
118
vl.c | 10 +-
119
docs/arm-cpu-features.rst | 37 ++++-
120
docs/devel/index.rst | 1 +
121
docs/devel/reset.rst | 289 +++++++++++++++++++++++++++++++++++++++
122
hw/core/trace-events | 27 ++++
123
51 files changed, 1727 insertions(+), 90 deletions(-)
124
create mode 100644 include/hw/resettable.h
125
create mode 100644 hw/core/resettable.c
126
create mode 100644 docs/devel/reset.rst
127
diff view generated by jsdifflib
1
The TypeInfo and state struct for bcm2386 disagree about what the
1
The num-lines property of the TYPE_OR_GATE device sets the number
2
parent class is -- the TypeInfo says it's TYPE_SYS_BUS_DEVICE,
2
of input lines it has. An assert() in or_irq_realize() restricts
3
but the BCM2386State struct only defines the parent_obj field
3
this to the maximum supported by the implementation. However we
4
as DeviceState. This would have caused problems if anything
4
got the condition in the assert wrong: it should be using <=,
5
actually tried to treat the object as a TYPE_SYS_BUS_DEVICE.
5
because num-lines == MAX_OR_LINES is permitted, and means that
6
Fix the TypeInfo to use TYPE_DEVICE as the parent, since we don't
6
all entries from 0 to MAX_OR_LINES-1 in the s->levels[] array
7
need any of the additional functionality TYPE_SYS_BUS_DEVICE
7
are used.
8
provides.
9
8
9
We didn't notice this previously because no user has so far
10
needed that many input lines.
11
12
Reported-by: Guenter Roeck <linux@roeck-us.net>
10
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
13
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
11
Reviewed-by: Andrew Baumann <Andrew.Baumann@microsoft.com>
14
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
12
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
15
Reviewed-by: Guenter Roeck <linux@roeck-us.net>
13
Message-id: 20180313153458.26822-5-peter.maydell@linaro.org
16
Message-id: 20200120142235.10432-1-peter.maydell@linaro.org
14
---
17
---
15
hw/arm/bcm2836.c | 2 +-
18
hw/core/or-irq.c | 2 +-
16
1 file changed, 1 insertion(+), 1 deletion(-)
19
1 file changed, 1 insertion(+), 1 deletion(-)
17
20
18
diff --git a/hw/arm/bcm2836.c b/hw/arm/bcm2836.c
21
diff --git a/hw/core/or-irq.c b/hw/core/or-irq.c
19
index XXXXXXX..XXXXXXX 100644
22
index XXXXXXX..XXXXXXX 100644
20
--- a/hw/arm/bcm2836.c
23
--- a/hw/core/or-irq.c
21
+++ b/hw/arm/bcm2836.c
24
+++ b/hw/core/or-irq.c
22
@@ -XXX,XX +XXX,XX @@ static void bcm2836_class_init(ObjectClass *oc, void *data)
25
@@ -XXX,XX +XXX,XX @@ static void or_irq_realize(DeviceState *dev, Error **errp)
23
26
{
24
static const TypeInfo bcm2836_type_info = {
27
qemu_or_irq *s = OR_IRQ(dev);
25
.name = TYPE_BCM2836,
28
26
- .parent = TYPE_SYS_BUS_DEVICE,
29
- assert(s->num_lines < MAX_OR_LINES);
27
+ .parent = TYPE_DEVICE,
30
+ assert(s->num_lines <= MAX_OR_LINES);
28
.instance_size = sizeof(BCM2836State),
31
29
.instance_init = bcm2836_init,
32
qdev_init_gpio_in(dev, or_irq_handler, s->num_lines);
30
.class_init = bcm2836_class_init,
33
}
31
--
34
--
32
2.16.2
35
2.20.1
33
36
34
37
diff view generated by jsdifflib
1
Our BCM2836 type is really a generic one that can be any of
1
The guest can use the semihosting API to open a handle
2
the bcm283x family. Rename it accordingly. We change only
2
corresponding to QEMU's own stdin, stdout, or stderr.
3
the names which are visible via the header file to the
3
When the guest closes this handle, we should not
4
rest of the QEMU code, leaving private function names
4
close the underlying host stdin/stdout/stderr
5
in bcm2836.c as they are.
5
the way we would do if the handle corresponded to
6
6
a host fd we'd opened on behalf of the guest in SYS_OPEN.
7
This is a preliminary to making bcm283x be an abstract
8
parent class to specific types for the bcm2836 and bcm2837.
9
7
10
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
8
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
11
Reviewed-by: Andrew Baumann <Andrew.Baumann@microsoft.com>
9
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
12
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
10
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
13
Message-id: 20180313153458.26822-6-peter.maydell@linaro.org
11
Message-id: 20200124172954.28481-1-peter.maydell@linaro.org
14
---
12
---
15
include/hw/arm/bcm2836.h | 12 ++++++------
13
target/arm/arm-semi.c | 9 +++++++++
16
hw/arm/bcm2836.c | 17 +++++++++--------
14
1 file changed, 9 insertions(+)
17
hw/arm/raspi.c | 16 ++++++++--------
18
3 files changed, 23 insertions(+), 22 deletions(-)
19
15
20
diff --git a/include/hw/arm/bcm2836.h b/include/hw/arm/bcm2836.h
16
diff --git a/target/arm/arm-semi.c b/target/arm/arm-semi.c
21
index XXXXXXX..XXXXXXX 100644
17
index XXXXXXX..XXXXXXX 100644
22
--- a/include/hw/arm/bcm2836.h
18
--- a/target/arm/arm-semi.c
23
+++ b/include/hw/arm/bcm2836.h
19
+++ b/target/arm/arm-semi.c
24
@@ -XXX,XX +XXX,XX @@
20
@@ -XXX,XX +XXX,XX @@ static uint32_t host_closefn(ARMCPU *cpu, GuestFD *gf)
25
#include "hw/arm/bcm2835_peripherals.h"
26
#include "hw/intc/bcm2836_control.h"
27
28
-#define TYPE_BCM2836 "bcm2836"
29
-#define BCM2836(obj) OBJECT_CHECK(BCM2836State, (obj), TYPE_BCM2836)
30
+#define TYPE_BCM283X "bcm283x"
31
+#define BCM283X(obj) OBJECT_CHECK(BCM283XState, (obj), TYPE_BCM283X)
32
33
-#define BCM2836_NCPUS 4
34
+#define BCM283X_NCPUS 4
35
36
-typedef struct BCM2836State {
37
+typedef struct BCM283XState {
38
/*< private >*/
39
DeviceState parent_obj;
40
/*< public >*/
41
@@ -XXX,XX +XXX,XX @@ typedef struct BCM2836State {
42
char *cpu_type;
43
uint32_t enabled_cpus;
44
45
- ARMCPU cpus[BCM2836_NCPUS];
46
+ ARMCPU cpus[BCM283X_NCPUS];
47
BCM2836ControlState control;
48
BCM2835PeripheralState peripherals;
49
-} BCM2836State;
50
+} BCM283XState;
51
52
#endif /* BCM2836_H */
53
diff --git a/hw/arm/bcm2836.c b/hw/arm/bcm2836.c
54
index XXXXXXX..XXXXXXX 100644
55
--- a/hw/arm/bcm2836.c
56
+++ b/hw/arm/bcm2836.c
57
@@ -XXX,XX +XXX,XX @@
58
59
static void bcm2836_init(Object *obj)
60
{
21
{
61
- BCM2836State *s = BCM2836(obj);
22
CPUARMState *env = &cpu->env;
62
+ BCM283XState *s = BCM283X(obj);
23
63
24
+ /*
64
object_initialize(&s->control, sizeof(s->control), TYPE_BCM2836_CONTROL);
25
+ * Only close the underlying host fd if it's one we opened on behalf
65
object_property_add_child(obj, "control", OBJECT(&s->control), NULL);
26
+ * of the guest in SYS_OPEN.
66
@@ -XXX,XX +XXX,XX @@ static void bcm2836_init(Object *obj)
27
+ */
67
28
+ if (gf->hostfd == STDIN_FILENO ||
68
static void bcm2836_realize(DeviceState *dev, Error **errp)
29
+ gf->hostfd == STDOUT_FILENO ||
69
{
30
+ gf->hostfd == STDERR_FILENO) {
70
- BCM2836State *s = BCM2836(dev);
31
+ return 0;
71
+ BCM283XState *s = BCM283X(dev);
32
+ }
72
Object *obj;
33
return set_swi_errno(env, close(gf->hostfd));
73
Error *err = NULL;
74
int n;
75
@@ -XXX,XX +XXX,XX @@ static void bcm2836_realize(DeviceState *dev, Error **errp)
76
/* common peripherals from bcm2835 */
77
78
obj = OBJECT(dev);
79
- for (n = 0; n < BCM2836_NCPUS; n++) {
80
+ for (n = 0; n < BCM283X_NCPUS; n++) {
81
object_initialize(&s->cpus[n], sizeof(s->cpus[n]),
82
s->cpu_type);
83
object_property_add_child(obj, "cpu[*]", OBJECT(&s->cpus[n]),
84
@@ -XXX,XX +XXX,XX @@ static void bcm2836_realize(DeviceState *dev, Error **errp)
85
sysbus_connect_irq(SYS_BUS_DEVICE(&s->peripherals), 1,
86
qdev_get_gpio_in_named(DEVICE(&s->control), "gpu-fiq", 0));
87
88
- for (n = 0; n < BCM2836_NCPUS; n++) {
89
+ for (n = 0; n < BCM283X_NCPUS; n++) {
90
/* Mirror bcm2836, which has clusterid set to 0xf
91
* TODO: this should be converted to a property of ARM_CPU
92
*/
93
@@ -XXX,XX +XXX,XX @@ static void bcm2836_realize(DeviceState *dev, Error **errp)
94
}
34
}
95
35
96
static Property bcm2836_props[] = {
97
- DEFINE_PROP_STRING("cpu-type", BCM2836State, cpu_type),
98
- DEFINE_PROP_UINT32("enabled-cpus", BCM2836State, enabled_cpus, BCM2836_NCPUS),
99
+ DEFINE_PROP_STRING("cpu-type", BCM283XState, cpu_type),
100
+ DEFINE_PROP_UINT32("enabled-cpus", BCM283XState, enabled_cpus,
101
+ BCM283X_NCPUS),
102
DEFINE_PROP_END_OF_LIST()
103
};
104
105
@@ -XXX,XX +XXX,XX @@ static void bcm2836_class_init(ObjectClass *oc, void *data)
106
}
107
108
static const TypeInfo bcm2836_type_info = {
109
- .name = TYPE_BCM2836,
110
+ .name = TYPE_BCM283X,
111
.parent = TYPE_DEVICE,
112
- .instance_size = sizeof(BCM2836State),
113
+ .instance_size = sizeof(BCM283XState),
114
.instance_init = bcm2836_init,
115
.class_init = bcm2836_class_init,
116
};
117
diff --git a/hw/arm/raspi.c b/hw/arm/raspi.c
118
index XXXXXXX..XXXXXXX 100644
119
--- a/hw/arm/raspi.c
120
+++ b/hw/arm/raspi.c
121
@@ -XXX,XX +XXX,XX @@
122
static const int raspi_boardid[] = {[1] = 0xc42, [2] = 0xc43, [3] = 0xc44};
123
124
typedef struct RasPiState {
125
- BCM2836State soc;
126
+ BCM283XState soc;
127
MemoryRegion ram;
128
} RasPiState;
129
130
@@ -XXX,XX +XXX,XX @@ static void raspi_init(MachineState *machine, int version)
131
BusState *bus;
132
DeviceState *carddev;
133
134
- object_initialize(&s->soc, sizeof(s->soc), TYPE_BCM2836);
135
+ object_initialize(&s->soc, sizeof(s->soc), TYPE_BCM283X);
136
object_property_add_child(OBJECT(machine), "soc", OBJECT(&s->soc),
137
&error_abort);
138
139
@@ -XXX,XX +XXX,XX @@ static void raspi2_machine_init(MachineClass *mc)
140
mc->no_floppy = 1;
141
mc->no_cdrom = 1;
142
mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
143
- mc->max_cpus = BCM2836_NCPUS;
144
- mc->min_cpus = BCM2836_NCPUS;
145
- mc->default_cpus = BCM2836_NCPUS;
146
+ mc->max_cpus = BCM283X_NCPUS;
147
+ mc->min_cpus = BCM283X_NCPUS;
148
+ mc->default_cpus = BCM283X_NCPUS;
149
mc->default_ram_size = 1024 * 1024 * 1024;
150
mc->ignore_memory_transaction_failures = true;
151
};
152
@@ -XXX,XX +XXX,XX @@ static void raspi3_machine_init(MachineClass *mc)
153
mc->no_floppy = 1;
154
mc->no_cdrom = 1;
155
mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a53");
156
- mc->max_cpus = BCM2836_NCPUS;
157
- mc->min_cpus = BCM2836_NCPUS;
158
- mc->default_cpus = BCM2836_NCPUS;
159
+ mc->max_cpus = BCM283X_NCPUS;
160
+ mc->min_cpus = BCM283X_NCPUS;
161
+ mc->default_cpus = BCM283X_NCPUS;
162
mc->default_ram_size = 1024 * 1024 * 1024;
163
}
164
DEFINE_MACHINE("raspi3", raspi3_machine_init)
165
--
36
--
166
2.16.2
37
2.20.1
167
38
168
39
diff view generated by jsdifflib
1
The BCM2837 sets the Aff1 field of the MPIDR affinity values for the
1
From: Andrew Jeffery <andrew@aj.id.au>
2
CPUs to 0, whereas the BCM2836 uses 0xf. Set this correctly, as it
3
is required for Linux to boot.
4
2
3
The AST2600 includes a second cut-down version of the SD/MMC controller
4
found in the AST2500, named the eMMC controller. It's cut down in the
5
sense that it only supports one slot rather than two, but it brings the
6
total number of slots supported by the AST2600 to three.
7
8
The existing code assumed that the SD controller always provided two
9
slots. Rework the SDHCI object to expose the number of slots as a
10
property to be set by the SoC configuration.
11
12
Signed-off-by: Andrew Jeffery <andrew@aj.id.au>
13
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
14
Reviewed-by: Cédric Le Goater <clg@kaod.org>
15
Signed-off-by: Cédric Le Goater <clg@kaod.org>
16
Message-id: 20200114103433.30534-2-clg@kaod.org
17
[PMM: fixed up to use device_class_set_props()]
5
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
18
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6
Reviewed-by: Andrew Baumann <Andrew.Baumann@microsoft.com>
7
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
8
Message-id: 20180313153458.26822-8-peter.maydell@linaro.org
9
---
19
---
10
hw/arm/bcm2836.c | 11 +++++++----
20
include/hw/sd/aspeed_sdhci.h | 1 +
11
1 file changed, 7 insertions(+), 4 deletions(-)
21
hw/arm/aspeed.c | 2 +-
22
hw/arm/aspeed_ast2600.c | 2 ++
23
hw/arm/aspeed_soc.c | 2 ++
24
hw/sd/aspeed_sdhci.c | 11 +++++++++--
25
5 files changed, 15 insertions(+), 3 deletions(-)
12
26
13
diff --git a/hw/arm/bcm2836.c b/hw/arm/bcm2836.c
27
diff --git a/include/hw/sd/aspeed_sdhci.h b/include/hw/sd/aspeed_sdhci.h
14
index XXXXXXX..XXXXXXX 100644
28
index XXXXXXX..XXXXXXX 100644
15
--- a/hw/arm/bcm2836.c
29
--- a/include/hw/sd/aspeed_sdhci.h
16
+++ b/hw/arm/bcm2836.c
30
+++ b/include/hw/sd/aspeed_sdhci.h
31
@@ -XXX,XX +XXX,XX @@ typedef struct AspeedSDHCIState {
32
SysBusDevice parent;
33
34
SDHCIState slots[ASPEED_SDHCI_NUM_SLOTS];
35
+ uint8_t num_slots;
36
37
MemoryRegion iomem;
38
qemu_irq irq;
39
diff --git a/hw/arm/aspeed.c b/hw/arm/aspeed.c
40
index XXXXXXX..XXXXXXX 100644
41
--- a/hw/arm/aspeed.c
42
+++ b/hw/arm/aspeed.c
43
@@ -XXX,XX +XXX,XX @@ static void aspeed_machine_init(MachineState *machine)
44
amc->i2c_init(bmc);
45
}
46
47
- for (i = 0; i < ARRAY_SIZE(bmc->soc.sdhci.slots); i++) {
48
+ for (i = 0; i < bmc->soc.sdhci.num_slots; i++) {
49
SDHCIState *sdhci = &bmc->soc.sdhci.slots[i];
50
DriveInfo *dinfo = drive_get_next(IF_SD);
51
BlockBackend *blk;
52
diff --git a/hw/arm/aspeed_ast2600.c b/hw/arm/aspeed_ast2600.c
53
index XXXXXXX..XXXXXXX 100644
54
--- a/hw/arm/aspeed_ast2600.c
55
+++ b/hw/arm/aspeed_ast2600.c
56
@@ -XXX,XX +XXX,XX @@ static void aspeed_soc_ast2600_init(Object *obj)
57
sysbus_init_child_obj(obj, "sdc", OBJECT(&s->sdhci), sizeof(s->sdhci),
58
TYPE_ASPEED_SDHCI);
59
60
+ object_property_set_int(OBJECT(&s->sdhci), 2, "num-slots", &error_abort);
61
+
62
/* Init sd card slot class here so that they're under the correct parent */
63
for (i = 0; i < ASPEED_SDHCI_NUM_SLOTS; ++i) {
64
sysbus_init_child_obj(obj, "sdhci[*]", OBJECT(&s->sdhci.slots[i]),
65
diff --git a/hw/arm/aspeed_soc.c b/hw/arm/aspeed_soc.c
66
index XXXXXXX..XXXXXXX 100644
67
--- a/hw/arm/aspeed_soc.c
68
+++ b/hw/arm/aspeed_soc.c
69
@@ -XXX,XX +XXX,XX @@ static void aspeed_soc_init(Object *obj)
70
sysbus_init_child_obj(obj, "sdc", OBJECT(&s->sdhci), sizeof(s->sdhci),
71
TYPE_ASPEED_SDHCI);
72
73
+ object_property_set_int(OBJECT(&s->sdhci), 2, "num-slots", &error_abort);
74
+
75
/* Init sd card slot class here so that they're under the correct parent */
76
for (i = 0; i < ASPEED_SDHCI_NUM_SLOTS; ++i) {
77
sysbus_init_child_obj(obj, "sdhci[*]", OBJECT(&s->sdhci.slots[i]),
78
diff --git a/hw/sd/aspeed_sdhci.c b/hw/sd/aspeed_sdhci.c
79
index XXXXXXX..XXXXXXX 100644
80
--- a/hw/sd/aspeed_sdhci.c
81
+++ b/hw/sd/aspeed_sdhci.c
17
@@ -XXX,XX +XXX,XX @@
82
@@ -XXX,XX +XXX,XX @@
18
83
#include "qapi/error.h"
19
struct BCM283XInfo {
84
#include "hw/irq.h"
20
const char *name;
85
#include "migration/vmstate.h"
21
+ int clusterid;
86
+#include "hw/qdev-properties.h"
22
};
87
23
88
#define ASPEED_SDHCI_INFO 0x00
24
static const BCM283XInfo bcm283x_socs[] = {
89
#define ASPEED_SDHCI_INFO_RESET 0x00030000
25
{
90
@@ -XXX,XX +XXX,XX @@ static void aspeed_sdhci_realize(DeviceState *dev, Error **errp)
26
.name = TYPE_BCM2836,
91
27
+ .clusterid = 0xf,
92
/* Create input irqs for the slots */
28
},
93
qdev_init_gpio_in_named_with_opaque(DEVICE(sbd), aspeed_sdhci_set_irq,
29
{
94
- sdhci, NULL, ASPEED_SDHCI_NUM_SLOTS);
30
.name = TYPE_BCM2837,
95
+ sdhci, NULL, sdhci->num_slots);
31
+ .clusterid = 0x0,
96
97
sysbus_init_irq(sbd, &sdhci->irq);
98
memory_region_init_io(&sdhci->iomem, OBJECT(sdhci), &aspeed_sdhci_ops,
99
sdhci, TYPE_ASPEED_SDHCI, 0x1000);
100
sysbus_init_mmio(sbd, &sdhci->iomem);
101
102
- for (int i = 0; i < ASPEED_SDHCI_NUM_SLOTS; ++i) {
103
+ for (int i = 0; i < sdhci->num_slots; ++i) {
104
Object *sdhci_slot = OBJECT(&sdhci->slots[i]);
105
SysBusDevice *sbd_slot = SYS_BUS_DEVICE(&sdhci->slots[i]);
106
107
@@ -XXX,XX +XXX,XX @@ static const VMStateDescription vmstate_aspeed_sdhci = {
32
},
108
},
33
};
109
};
34
110
35
@@ -XXX,XX +XXX,XX @@ static void bcm2836_init(Object *obj)
111
+static Property aspeed_sdhci_properties[] = {
36
static void bcm2836_realize(DeviceState *dev, Error **errp)
112
+ DEFINE_PROP_UINT8("num-slots", AspeedSDHCIState, num_slots, 0),
113
+ DEFINE_PROP_END_OF_LIST(),
114
+};
115
+
116
static void aspeed_sdhci_class_init(ObjectClass *classp, void *data)
37
{
117
{
38
BCM283XState *s = BCM283X(dev);
118
DeviceClass *dc = DEVICE_CLASS(classp);
39
+ BCM283XClass *bc = BCM283X_GET_CLASS(dev);
119
@@ -XXX,XX +XXX,XX @@ static void aspeed_sdhci_class_init(ObjectClass *classp, void *data)
40
+ const BCM283XInfo *info = bc->info;
120
dc->realize = aspeed_sdhci_realize;
41
Object *obj;
121
dc->reset = aspeed_sdhci_reset;
42
Error *err = NULL;
122
dc->vmsd = &vmstate_aspeed_sdhci;
43
int n;
123
+ device_class_set_props(dc, aspeed_sdhci_properties);
44
@@ -XXX,XX +XXX,XX @@ static void bcm2836_realize(DeviceState *dev, Error **errp)
124
}
45
qdev_get_gpio_in_named(DEVICE(&s->control), "gpu-fiq", 0));
125
46
126
static TypeInfo aspeed_sdhci_info = {
47
for (n = 0; n < BCM283X_NCPUS; n++) {
48
- /* Mirror bcm2836, which has clusterid set to 0xf
49
- * TODO: this should be converted to a property of ARM_CPU
50
- */
51
- s->cpus[n].mp_affinity = 0xF00 | n;
52
+ /* TODO: this should be converted to a property of ARM_CPU */
53
+ s->cpus[n].mp_affinity = (info->clusterid << 8) | n;
54
55
/* set periphbase/CBAR value for CPU-local registers */
56
object_property_set_int(OBJECT(&s->cpus[n]),
57
--
127
--
58
2.16.2
128
2.20.1
59
129
60
130
diff view generated by jsdifflib
1
The raspi3 has AArch64 CPUs, which means that our smpboot
1
From: Andrew Jeffery <andrew@aj.id.au>
2
code for keeping the secondary CPUs in a pen needs to have
3
a version for A64 as well as A32. Without this, the
4
secondary CPUs go into an infinite loop of taking undefined
5
instruction exceptions.
6
2
3
Initialise another SDHCI model instance for the AST2600's eMMC
4
controller and use the SDHCI's num_slots value introduced previously to
5
determine whether we should create an SD card instance for the new slot.
6
7
Signed-off-by: Andrew Jeffery <andrew@aj.id.au>
8
Reviewed-by: Cédric Le Goater <clg@kaod.org>
9
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
10
Signed-off-by: Cédric Le Goater <clg@kaod.org>
11
Message-id: 20200114103433.30534-3-clg@kaod.org
12
[ clg : - removed ternary operator from sdhci_attach_drive()
13
- renamed SDHCI objects with a '-controller' prefix ]
14
Signed-off-by: Cédric Le Goater <clg@kaod.org>
7
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
15
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
8
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
9
Message-id: 20180313153458.26822-10-peter.maydell@linaro.org
10
---
16
---
11
hw/arm/raspi.c | 41 ++++++++++++++++++++++++++++++++++++++++-
17
include/hw/arm/aspeed_soc.h | 2 ++
12
1 file changed, 40 insertions(+), 1 deletion(-)
18
hw/arm/aspeed.c | 26 +++++++++++++++++---------
19
hw/arm/aspeed_ast2600.c | 29 ++++++++++++++++++++++++++---
20
3 files changed, 45 insertions(+), 12 deletions(-)
13
21
14
diff --git a/hw/arm/raspi.c b/hw/arm/raspi.c
22
diff --git a/include/hw/arm/aspeed_soc.h b/include/hw/arm/aspeed_soc.h
15
index XXXXXXX..XXXXXXX 100644
23
index XXXXXXX..XXXXXXX 100644
16
--- a/hw/arm/raspi.c
24
--- a/include/hw/arm/aspeed_soc.h
17
+++ b/hw/arm/raspi.c
25
+++ b/include/hw/arm/aspeed_soc.h
18
@@ -XXX,XX +XXX,XX @@
26
@@ -XXX,XX +XXX,XX @@ typedef struct AspeedSoCState {
19
#define BOARDSETUP_ADDR (MVBAR_ADDR + 0x20) /* board setup code */
27
AspeedGPIOState gpio;
20
#define FIRMWARE_ADDR_2 0x8000 /* Pi 2 loads kernel.img here by default */
28
AspeedGPIOState gpio_1_8v;
21
#define FIRMWARE_ADDR_3 0x80000 /* Pi 3 loads kernel.img here by default */
29
AspeedSDHCIState sdhci;
22
+#define SPINTABLE_ADDR 0xd8 /* Pi 3 bootloader spintable */
30
+ AspeedSDHCIState emmc;
23
31
} AspeedSoCState;
24
/* Table of Linux board IDs for different Pi versions */
32
25
static const int raspi_boardid[] = {[1] = 0xc42, [2] = 0xc43, [3] = 0xc44};
33
#define TYPE_ASPEED_SOC "aspeed-soc"
26
@@ -XXX,XX +XXX,XX @@ static void write_smpboot(ARMCPU *cpu, const struct arm_boot_info *info)
34
@@ -XXX,XX +XXX,XX @@ enum {
27
info->smp_loader_start);
35
ASPEED_MII4,
36
ASPEED_SDRAM,
37
ASPEED_XDMA,
38
+ ASPEED_EMMC,
39
};
40
41
#endif /* ASPEED_SOC_H */
42
diff --git a/hw/arm/aspeed.c b/hw/arm/aspeed.c
43
index XXXXXXX..XXXXXXX 100644
44
--- a/hw/arm/aspeed.c
45
+++ b/hw/arm/aspeed.c
46
@@ -XXX,XX +XXX,XX @@ static void aspeed_board_init_flashes(AspeedSMCState *s, const char *flashtype,
47
}
28
}
48
}
29
49
30
+static void write_smpboot64(ARMCPU *cpu, const struct arm_boot_info *info)
50
+static void sdhci_attach_drive(SDHCIState *sdhci, DriveInfo *dinfo)
31
+{
51
+{
32
+ /* Unlike the AArch32 version we don't need to call the board setup hook.
52
+ DeviceState *card;
33
+ * The mechanism for doing the spin-table is also entirely different.
34
+ * We must have four 64-bit fields at absolute addresses
35
+ * 0xd8, 0xe0, 0xe8, 0xf0 in RAM, which are the flag variables for
36
+ * our CPUs, and which we must ensure are zero initialized before
37
+ * the primary CPU goes into the kernel. We put these variables inside
38
+ * a rom blob, so that the reset for ROM contents zeroes them for us.
39
+ */
40
+ static const uint32_t smpboot[] = {
41
+ 0xd2801b05, /* mov x5, 0xd8 */
42
+ 0xd53800a6, /* mrs x6, mpidr_el1 */
43
+ 0x924004c6, /* and x6, x6, #0x3 */
44
+ 0xd503205f, /* spin: wfe */
45
+ 0xf86678a4, /* ldr x4, [x5,x6,lsl #3] */
46
+ 0xb4ffffc4, /* cbz x4, spin */
47
+ 0xd2800000, /* mov x0, #0x0 */
48
+ 0xd2800001, /* mov x1, #0x0 */
49
+ 0xd2800002, /* mov x2, #0x0 */
50
+ 0xd2800003, /* mov x3, #0x0 */
51
+ 0xd61f0080, /* br x4 */
52
+ };
53
+
53
+
54
+ static const uint64_t spintables[] = {
54
+ card = qdev_create(qdev_get_child_bus(DEVICE(sdhci), "sd-bus"),
55
+ 0, 0, 0, 0
55
+ TYPE_SD_CARD);
56
+ };
56
+ if (dinfo) {
57
+
57
+ qdev_prop_set_drive(card, "drive", blk_by_legacy_dinfo(dinfo),
58
+ rom_add_blob_fixed("raspi_smpboot", smpboot, sizeof(smpboot),
58
+ &error_fatal);
59
+ info->smp_loader_start);
59
+ }
60
+ rom_add_blob_fixed("raspi_spintables", spintables, sizeof(spintables),
60
+ object_property_set_bool(OBJECT(card), true, "realized", &error_fatal);
61
+ SPINTABLE_ADDR);
62
+}
61
+}
63
+
62
+
64
static void write_board_setup(ARMCPU *cpu, const struct arm_boot_info *info)
63
static void aspeed_machine_init(MachineState *machine)
65
{
64
{
66
arm_write_secure_board_setup_dummy_smc(cpu, info, MVBAR_ADDR);
65
AspeedBoardState *bmc;
67
@@ -XXX,XX +XXX,XX @@ static void setup_boot(MachineState *machine, int version, size_t ram_size)
66
@@ -XXX,XX +XXX,XX @@ static void aspeed_machine_init(MachineState *machine)
68
/* Pi2 and Pi3 requires SMP setup */
69
if (version >= 2) {
70
binfo.smp_loader_start = SMPBOOT_ADDR;
71
- binfo.write_secondary_boot = write_smpboot;
72
+ if (version == 2) {
73
+ binfo.write_secondary_boot = write_smpboot;
74
+ } else {
75
+ binfo.write_secondary_boot = write_smpboot64;
76
+ }
77
binfo.secondary_cpu_reset_hook = reset_secondary;
78
}
67
}
79
68
69
for (i = 0; i < bmc->soc.sdhci.num_slots; i++) {
70
- SDHCIState *sdhci = &bmc->soc.sdhci.slots[i];
71
- DriveInfo *dinfo = drive_get_next(IF_SD);
72
- BlockBackend *blk;
73
- DeviceState *card;
74
+ sdhci_attach_drive(&bmc->soc.sdhci.slots[i], drive_get_next(IF_SD));
75
+ }
76
77
- blk = dinfo ? blk_by_legacy_dinfo(dinfo) : NULL;
78
- card = qdev_create(qdev_get_child_bus(DEVICE(sdhci), "sd-bus"),
79
- TYPE_SD_CARD);
80
- qdev_prop_set_drive(card, "drive", blk, &error_fatal);
81
- object_property_set_bool(OBJECT(card), true, "realized", &error_fatal);
82
+ if (bmc->soc.emmc.num_slots) {
83
+ sdhci_attach_drive(&bmc->soc.emmc.slots[0], drive_get_next(IF_SD));
84
}
85
86
arm_load_kernel(ARM_CPU(first_cpu), machine, &aspeed_board_binfo);
87
diff --git a/hw/arm/aspeed_ast2600.c b/hw/arm/aspeed_ast2600.c
88
index XXXXXXX..XXXXXXX 100644
89
--- a/hw/arm/aspeed_ast2600.c
90
+++ b/hw/arm/aspeed_ast2600.c
91
@@ -XXX,XX +XXX,XX @@ static const hwaddr aspeed_soc_ast2600_memmap[] = {
92
[ASPEED_ADC] = 0x1E6E9000,
93
[ASPEED_VIDEO] = 0x1E700000,
94
[ASPEED_SDHCI] = 0x1E740000,
95
+ [ASPEED_EMMC] = 0x1E750000,
96
[ASPEED_GPIO] = 0x1E780000,
97
[ASPEED_GPIO_1_8V] = 0x1E780800,
98
[ASPEED_RTC] = 0x1E781000,
99
@@ -XXX,XX +XXX,XX @@ static const hwaddr aspeed_soc_ast2600_memmap[] = {
100
101
#define ASPEED_SOC_AST2600_MAX_IRQ 128
102
103
+/* Shared Peripheral Interrupt values below are offset by -32 from datasheet */
104
static const int aspeed_soc_ast2600_irqmap[] = {
105
[ASPEED_UART1] = 47,
106
[ASPEED_UART2] = 48,
107
@@ -XXX,XX +XXX,XX @@ static const int aspeed_soc_ast2600_irqmap[] = {
108
[ASPEED_ADC] = 78,
109
[ASPEED_XDMA] = 6,
110
[ASPEED_SDHCI] = 43,
111
+ [ASPEED_EMMC] = 15,
112
[ASPEED_GPIO] = 40,
113
[ASPEED_GPIO_1_8V] = 11,
114
[ASPEED_RTC] = 13,
115
@@ -XXX,XX +XXX,XX @@ static void aspeed_soc_ast2600_init(Object *obj)
116
sysbus_init_child_obj(obj, "gpio_1_8v", OBJECT(&s->gpio_1_8v),
117
sizeof(s->gpio_1_8v), typename);
118
119
- sysbus_init_child_obj(obj, "sdc", OBJECT(&s->sdhci), sizeof(s->sdhci),
120
- TYPE_ASPEED_SDHCI);
121
+ sysbus_init_child_obj(obj, "sd-controller", OBJECT(&s->sdhci),
122
+ sizeof(s->sdhci), TYPE_ASPEED_SDHCI);
123
124
object_property_set_int(OBJECT(&s->sdhci), 2, "num-slots", &error_abort);
125
126
/* Init sd card slot class here so that they're under the correct parent */
127
for (i = 0; i < ASPEED_SDHCI_NUM_SLOTS; ++i) {
128
- sysbus_init_child_obj(obj, "sdhci[*]", OBJECT(&s->sdhci.slots[i]),
129
+ sysbus_init_child_obj(obj, "sd-controller.sdhci[*]",
130
+ OBJECT(&s->sdhci.slots[i]),
131
sizeof(s->sdhci.slots[i]), TYPE_SYSBUS_SDHCI);
132
}
133
+
134
+ sysbus_init_child_obj(obj, "emmc-controller", OBJECT(&s->emmc),
135
+ sizeof(s->emmc), TYPE_ASPEED_SDHCI);
136
+
137
+ object_property_set_int(OBJECT(&s->emmc), 1, "num-slots", &error_abort);
138
+
139
+ sysbus_init_child_obj(obj, "emmc-controller.sdhci",
140
+ OBJECT(&s->emmc.slots[0]), sizeof(s->emmc.slots[0]),
141
+ TYPE_SYSBUS_SDHCI);
142
}
143
144
/*
145
@@ -XXX,XX +XXX,XX @@ static void aspeed_soc_ast2600_realize(DeviceState *dev, Error **errp)
146
sc->memmap[ASPEED_SDHCI]);
147
sysbus_connect_irq(SYS_BUS_DEVICE(&s->sdhci), 0,
148
aspeed_soc_get_irq(s, ASPEED_SDHCI));
149
+
150
+ /* eMMC */
151
+ object_property_set_bool(OBJECT(&s->emmc), true, "realized", &err);
152
+ if (err) {
153
+ error_propagate(errp, err);
154
+ return;
155
+ }
156
+ sysbus_mmio_map(SYS_BUS_DEVICE(&s->emmc), 0, sc->memmap[ASPEED_EMMC]);
157
+ sysbus_connect_irq(SYS_BUS_DEVICE(&s->emmc), 0,
158
+ aspeed_soc_get_irq(s, ASPEED_EMMC));
159
}
160
161
static void aspeed_soc_ast2600_class_init(ObjectClass *oc, void *data)
80
--
162
--
81
2.16.2
163
2.20.1
82
164
83
165
diff view generated by jsdifflib
1
From: Andrey Smirnov <andrew.smirnov@gmail.com>
1
From: Cédric Le Goater <clg@kaod.org>
2
2
3
Add support for "TX complete"/TXDC interrupt generate by real HW since
3
These buffers should be aligned on 16 bytes.
4
it is needed to support guests other than Linux.
5
4
6
Based on the patch by Bill Paul as found here:
5
Ignore invalid RX and TX buffer addresses and log an error. All
7
https://bugs.launchpad.net/qemu/+bug/1753314
6
incoming and outgoing traffic will be dropped because no valid RX or
7
TX descriptors will be available.
8
8
9
Cc: qemu-devel@nongnu.org
9
Signed-off-by: Cédric Le Goater <clg@kaod.org>
10
Cc: qemu-arm@nongnu.org
10
Message-id: 20200114103433.30534-4-clg@kaod.org
11
Cc: Bill Paul <wpaul@windriver.com>
12
Cc: Peter Maydell <peter.maydell@linaro.org>
13
Signed-off-by: Bill Paul <wpaul@windriver.com>
14
Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
15
Message-id: 20180315191141.6789-2-andrew.smirnov@gmail.com
16
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
11
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
17
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
12
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
18
---
13
---
19
include/hw/char/imx_serial.h | 3 +++
14
hw/net/ftgmac100.c | 13 +++++++++++++
20
hw/char/imx_serial.c | 20 +++++++++++++++++---
15
1 file changed, 13 insertions(+)
21
2 files changed, 20 insertions(+), 3 deletions(-)
22
16
23
diff --git a/include/hw/char/imx_serial.h b/include/hw/char/imx_serial.h
17
diff --git a/hw/net/ftgmac100.c b/hw/net/ftgmac100.c
24
index XXXXXXX..XXXXXXX 100644
18
index XXXXXXX..XXXXXXX 100644
25
--- a/include/hw/char/imx_serial.h
19
--- a/hw/net/ftgmac100.c
26
+++ b/include/hw/char/imx_serial.h
20
+++ b/hw/net/ftgmac100.c
27
@@ -XXX,XX +XXX,XX @@
21
@@ -XXX,XX +XXX,XX @@ typedef struct {
28
#define UCR2_RXEN (1<<1) /* Receiver enable */
22
uint32_t des3;
29
#define UCR2_SRST (1<<0) /* Reset complete */
23
} FTGMAC100Desc;
30
24
31
+#define UCR4_TCEN BIT(3) /* TX complete interrupt enable */
25
+#define FTGMAC100_DESC_ALIGNMENT 16
32
+
26
+
33
#define UTS1_TXEMPTY (1<<6)
27
/*
34
#define UTS1_RXEMPTY (1<<5)
28
* Specific RTL8211E MII Registers
35
#define UTS1_TXFULL (1<<4)
29
*/
36
@@ -XXX,XX +XXX,XX @@ typedef struct IMXSerialState {
30
@@ -XXX,XX +XXX,XX @@ static void ftgmac100_write(void *opaque, hwaddr addr,
37
uint32_t ubmr;
31
s->itc = value;
38
uint32_t ubrc;
32
break;
39
uint32_t ucr3;
33
case FTGMAC100_RXR_BADR: /* Ring buffer address */
40
+ uint32_t ucr4;
34
+ if (!QEMU_IS_ALIGNED(value, FTGMAC100_DESC_ALIGNMENT)) {
41
35
+ qemu_log_mask(LOG_GUEST_ERROR, "%s: Bad RX buffer alignment 0x%"
42
qemu_irq irq;
36
+ HWADDR_PRIx "\n", __func__, value);
43
CharBackend chr;
37
+ return;
44
diff --git a/hw/char/imx_serial.c b/hw/char/imx_serial.c
38
+ }
45
index XXXXXXX..XXXXXXX 100644
46
--- a/hw/char/imx_serial.c
47
+++ b/hw/char/imx_serial.c
48
@@ -XXX,XX +XXX,XX @@
49
50
static const VMStateDescription vmstate_imx_serial = {
51
.name = TYPE_IMX_SERIAL,
52
- .version_id = 1,
53
- .minimum_version_id = 1,
54
+ .version_id = 2,
55
+ .minimum_version_id = 2,
56
.fields = (VMStateField[]) {
57
VMSTATE_INT32(readbuff, IMXSerialState),
58
VMSTATE_UINT32(usr1, IMXSerialState),
59
@@ -XXX,XX +XXX,XX @@ static const VMStateDescription vmstate_imx_serial = {
60
VMSTATE_UINT32(ubmr, IMXSerialState),
61
VMSTATE_UINT32(ubrc, IMXSerialState),
62
VMSTATE_UINT32(ucr3, IMXSerialState),
63
+ VMSTATE_UINT32(ucr4, IMXSerialState),
64
VMSTATE_END_OF_LIST()
65
},
66
};
67
@@ -XXX,XX +XXX,XX @@ static void imx_update(IMXSerialState *s)
68
* unfortunately.
69
*/
70
mask = (s->ucr1 & UCR1_TXMPTYEN) ? USR2_TXFE : 0;
71
+ /*
72
+ * TCEN and TXDC are both bit 3
73
+ */
74
+ mask |= s->ucr4 & UCR4_TCEN;
75
+
39
+
76
usr2 = s->usr2 & mask;
40
s->rx_ring = value;
77
41
s->rx_descriptor = s->rx_ring;
78
qemu_set_irq(s->irq, usr1 || usr2);
79
@@ -XXX,XX +XXX,XX @@ static uint64_t imx_serial_read(void *opaque, hwaddr offset,
80
return s->ucr3;
81
82
case 0x23: /* UCR4 */
83
+ return s->ucr4;
84
+
85
case 0x29: /* BRM Incremental */
86
return 0x0; /* TODO */
87
88
@@ -XXX,XX +XXX,XX @@ static void imx_serial_write(void *opaque, hwaddr offset,
89
* qemu_chr_fe_write and background I/O callbacks */
90
qemu_chr_fe_write_all(&s->chr, &ch, 1);
91
s->usr1 &= ~USR1_TRDY;
92
+ s->usr2 &= ~USR2_TXDC;
93
imx_update(s);
94
s->usr1 |= USR1_TRDY;
95
+ s->usr2 |= USR2_TXDC;
96
imx_update(s);
97
}
98
break;
42
break;
99
@@ -XXX,XX +XXX,XX @@ static void imx_serial_write(void *opaque, hwaddr offset,
43
@@ -XXX,XX +XXX,XX @@ static void ftgmac100_write(void *opaque, hwaddr addr,
100
s->ucr3 = value & 0xffff;
101
break;
44
break;
102
45
103
- case 0x2d: /* UTS1 */
46
case FTGMAC100_NPTXR_BADR: /* Transmit buffer address */
104
case 0x23: /* UCR4 */
47
+ if (!QEMU_IS_ALIGNED(value, FTGMAC100_DESC_ALIGNMENT)) {
105
+ s->ucr4 = value & 0xffff;
48
+ qemu_log_mask(LOG_GUEST_ERROR, "%s: Bad TX buffer alignment 0x%"
106
+ imx_update(s);
49
+ HWADDR_PRIx "\n", __func__, value);
107
+ break;
50
+ return;
108
+
51
+ }
109
+ case 0x2d: /* UTS1 */
52
s->tx_ring = value;
110
qemu_log_mask(LOG_UNIMP, "[%s]%s: Unimplemented reg 0x%"
53
s->tx_descriptor = s->tx_ring;
111
HWADDR_PRIx "\n", TYPE_IMX_SERIAL, __func__, offset);
54
break;
112
/* TODO */
113
--
55
--
114
2.16.2
56
2.20.1
115
57
116
58
diff view generated by jsdifflib
New patch
1
From: Cédric Le Goater <clg@kaod.org>
1
2
3
The overhead for the OpenBMC firmware images using the a custom U-Boot
4
is around 2 seconds, which is fine, but with a U-Boot from mainline,
5
it takes an extra 50 seconds or so to reach Linux. A quick survey on
6
the number of reads performed on the flash memory region gives the
7
following figures :
8
9
OpenBMC U-Boot 922478 (~ 3.5 MBytes)
10
Mainline U-Boot 20569977 (~ 80 MBytes)
11
12
QEMU must be trashing the TCG TBs and reloading text very often. Some
13
addresses are read more than 250.000 times. Until we find a solution
14
to improve boot time, execution from MMIO is not activated by default.
15
16
Setting this option also breaks migration compatibility.
17
18
Signed-off-by: Cédric Le Goater <clg@kaod.org>
19
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
20
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
21
Message-id: 20200114103433.30534-5-clg@kaod.org
22
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
23
---
24
include/hw/arm/aspeed.h | 2 ++
25
hw/arm/aspeed.c | 44 ++++++++++++++++++++++++++++++++++++-----
26
2 files changed, 41 insertions(+), 5 deletions(-)
27
28
diff --git a/include/hw/arm/aspeed.h b/include/hw/arm/aspeed.h
29
index XXXXXXX..XXXXXXX 100644
30
--- a/include/hw/arm/aspeed.h
31
+++ b/include/hw/arm/aspeed.h
32
@@ -XXX,XX +XXX,XX @@ typedef struct AspeedBoardState AspeedBoardState;
33
34
typedef struct AspeedMachine {
35
MachineState parent_obj;
36
+
37
+ bool mmio_exec;
38
} AspeedMachine;
39
40
#define ASPEED_MACHINE_CLASS(klass) \
41
diff --git a/hw/arm/aspeed.c b/hw/arm/aspeed.c
42
index XXXXXXX..XXXXXXX 100644
43
--- a/hw/arm/aspeed.c
44
+++ b/hw/arm/aspeed.c
45
@@ -XXX,XX +XXX,XX @@ static void aspeed_machine_init(MachineState *machine)
46
* SoC and 128MB for the AST2500 SoC, which is twice as big as
47
* needed by the flash modules of the Aspeed machines.
48
*/
49
- memory_region_init_rom(boot_rom, OBJECT(bmc), "aspeed.boot_rom",
50
- fl->size, &error_abort);
51
- memory_region_add_subregion(get_system_memory(), FIRMWARE_ADDR,
52
- boot_rom);
53
- write_boot_rom(drive0, FIRMWARE_ADDR, fl->size, &error_abort);
54
+ if (ASPEED_MACHINE(machine)->mmio_exec) {
55
+ memory_region_init_alias(boot_rom, OBJECT(bmc), "aspeed.boot_rom",
56
+ &fl->mmio, 0, fl->size);
57
+ memory_region_add_subregion(get_system_memory(), FIRMWARE_ADDR,
58
+ boot_rom);
59
+ } else {
60
+ memory_region_init_rom(boot_rom, OBJECT(bmc), "aspeed.boot_rom",
61
+ fl->size, &error_abort);
62
+ memory_region_add_subregion(get_system_memory(), FIRMWARE_ADDR,
63
+ boot_rom);
64
+ write_boot_rom(drive0, FIRMWARE_ADDR, fl->size, &error_abort);
65
+ }
66
}
67
68
aspeed_board_binfo.ram_size = ram_size;
69
@@ -XXX,XX +XXX,XX @@ static void witherspoon_bmc_i2c_init(AspeedBoardState *bmc)
70
/* Bus 11: TODO ucd90160@64 */
71
}
72
73
+static bool aspeed_get_mmio_exec(Object *obj, Error **errp)
74
+{
75
+ return ASPEED_MACHINE(obj)->mmio_exec;
76
+}
77
+
78
+static void aspeed_set_mmio_exec(Object *obj, bool value, Error **errp)
79
+{
80
+ ASPEED_MACHINE(obj)->mmio_exec = value;
81
+}
82
+
83
+static void aspeed_machine_instance_init(Object *obj)
84
+{
85
+ ASPEED_MACHINE(obj)->mmio_exec = false;
86
+}
87
+
88
+static void aspeed_machine_class_props_init(ObjectClass *oc)
89
+{
90
+ object_class_property_add_bool(oc, "execute-in-place",
91
+ aspeed_get_mmio_exec,
92
+ aspeed_set_mmio_exec, &error_abort);
93
+ object_class_property_set_description(oc, "execute-in-place",
94
+ "boot directly from CE0 flash device", &error_abort);
95
+}
96
+
97
static void aspeed_machine_class_init(ObjectClass *oc, void *data)
98
{
99
MachineClass *mc = MACHINE_CLASS(oc);
100
@@ -XXX,XX +XXX,XX @@ static void aspeed_machine_class_init(ObjectClass *oc, void *data)
101
mc->no_floppy = 1;
102
mc->no_cdrom = 1;
103
mc->no_parallel = 1;
104
+
105
+ aspeed_machine_class_props_init(oc);
106
}
107
108
static void aspeed_machine_palmetto_class_init(ObjectClass *oc, void *data)
109
@@ -XXX,XX +XXX,XX @@ static const TypeInfo aspeed_machine_types[] = {
110
.name = TYPE_ASPEED_MACHINE,
111
.parent = TYPE_MACHINE,
112
.instance_size = sizeof(AspeedMachine),
113
+ .instance_init = aspeed_machine_instance_init,
114
.class_size = sizeof(AspeedMachineClass),
115
.class_init = aspeed_machine_class_init,
116
.abstract = true,
117
--
118
2.20.1
119
120
diff view generated by jsdifflib
1
From: Andrey Smirnov <andrew.smirnov@gmail.com>
1
From: Joel Stanley <joel@jms.id.au>
2
2
3
Code of imx_update() is slightly confusing since the "flags" variable
3
Following the pattern of the work recently done with the ASPEED GPIO
4
doesn't really corespond to anything in real hardware and server as a
4
model, this adds support for inspecting and modifying the PCA9552 LEDs
5
kitchensink accumulating events normally reported via USR1 and USR2
5
from the monitor.
6
registers.
7
6
8
Change the code to explicitly evaluate state of interrupts reported
7
(qemu) qom-set /machine/unattached/device[17] led0 on
9
via USR1 and USR2 against corresponding masking bits and use the to
8
(qemu) qom-set /machine/unattached/device[17] led0 off
10
detemine if IRQ line should be asserted or not.
9
(qemu) qom-set /machine/unattached/device[17] led0 pwm0
10
(qemu) qom-set /machine/unattached/device[17] led0 pwm1
11
11
12
NOTE: Check for UTS1_TXEMPTY being set has been dropped for two
12
Signed-off-by: Joel Stanley <joel@jms.id.au>
13
reasons:
13
Signed-off-by: Cédric Le Goater <clg@kaod.org>
14
14
Message-id: 20200114103433.30534-6-clg@kaod.org
15
1. Emulation code implements a single character FIFO, so this flag
15
[clg: - removed the "qom-get" examples from the commit log
16
will always be set since characters are trasmitted as a part of
16
- merged memory leak fixes from Joel ]
17
the code emulating "push" into the FIFO
17
Signed-off-by: Cédric Le Goater <clg@kaod.org>
18
19
2. imx_update() is really just a function doing ORing and maksing
20
of reported events, so checking for UTS1_TXEMPTY should happen,
21
if it's ever really needed should probably happen outside of
22
it.
23
24
Cc: qemu-devel@nongnu.org
25
Cc: qemu-arm@nongnu.org
26
Cc: Bill Paul <wpaul@windriver.com>
27
Cc: Peter Maydell <peter.maydell@linaro.org>
28
Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
29
Message-id: 20180315191141.6789-1-andrew.smirnov@gmail.com
30
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
18
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
31
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
19
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
32
---
20
---
33
hw/char/imx_serial.c | 24 ++++++++++++++++--------
21
hw/misc/pca9552.c | 90 +++++++++++++++++++++++++++++++++++++++++++++++
34
1 file changed, 16 insertions(+), 8 deletions(-)
22
1 file changed, 90 insertions(+)
35
23
36
diff --git a/hw/char/imx_serial.c b/hw/char/imx_serial.c
24
diff --git a/hw/misc/pca9552.c b/hw/misc/pca9552.c
37
index XXXXXXX..XXXXXXX 100644
25
index XXXXXXX..XXXXXXX 100644
38
--- a/hw/char/imx_serial.c
26
--- a/hw/misc/pca9552.c
39
+++ b/hw/char/imx_serial.c
27
+++ b/hw/misc/pca9552.c
40
@@ -XXX,XX +XXX,XX @@ static const VMStateDescription vmstate_imx_serial = {
28
@@ -XXX,XX +XXX,XX @@
41
29
#include "hw/misc/pca9552.h"
42
static void imx_update(IMXSerialState *s)
30
#include "hw/misc/pca9552_regs.h"
31
#include "migration/vmstate.h"
32
+#include "qapi/error.h"
33
+#include "qapi/visitor.h"
34
35
#define PCA9552_LED_ON 0x0
36
#define PCA9552_LED_OFF 0x1
37
#define PCA9552_LED_PWM0 0x2
38
#define PCA9552_LED_PWM1 0x3
39
40
+static const char *led_state[] = {"on", "off", "pwm0", "pwm1"};
41
+
42
static uint8_t pca9552_pin_get_config(PCA9552State *s, int pin)
43
{
43
{
44
- uint32_t flags;
44
uint8_t reg = PCA9552_LS0 + (pin / 4);
45
+ uint32_t usr1;
45
@@ -XXX,XX +XXX,XX @@ static int pca9552_event(I2CSlave *i2c, enum i2c_event event)
46
+ uint32_t usr2;
46
return 0;
47
+ uint32_t mask;
47
}
48
48
49
- flags = (s->usr1 & s->ucr1) & (USR1_TRDY|USR1_RRDY);
49
+static void pca9552_get_led(Object *obj, Visitor *v, const char *name,
50
- if (s->ucr1 & UCR1_TXMPTYEN) {
50
+ void *opaque, Error **errp)
51
- flags |= (s->uts1 & UTS1_TXEMPTY);
51
+{
52
- } else {
52
+ PCA9552State *s = PCA9552(obj);
53
- flags &= ~USR1_TRDY;
53
+ int led, rc, reg;
54
- }
54
+ uint8_t state;
55
+
56
+ rc = sscanf(name, "led%2d", &led);
57
+ if (rc != 1) {
58
+ error_setg(errp, "%s: error reading %s", __func__, name);
59
+ return;
60
+ }
61
+ if (led < 0 || led > s->nr_leds) {
62
+ error_setg(errp, "%s invalid led %s", __func__, name);
63
+ return;
64
+ }
55
+ /*
65
+ /*
56
+ * Lucky for us TRDY and RRDY has the same offset in both USR1 and
66
+ * Get the LSx register as the qom interface should expose the device
57
+ * UCR1, so we can get away with something as simple as the
67
+ * state, not the modeled 'input line' behaviour which would come from
58
+ * following:
68
+ * reading the INPUTx reg
59
+ */
69
+ */
60
+ usr1 = s->usr1 & s->ucr1 & (USR1_TRDY | USR1_RRDY);
70
+ reg = PCA9552_LS0 + led / 4;
61
+ /*
71
+ state = (pca9552_read(s, reg) >> (led % 8)) & 0x3;
62
+ * Bits that we want in USR2 are not as conveniently laid out,
72
+ visit_type_str(v, name, (char **)&led_state[state], errp);
63
+ * unfortunately.
73
+}
64
+ */
74
+
65
+ mask = (s->ucr1 & UCR1_TXMPTYEN) ? USR2_TXFE : 0;
75
+/*
66
+ usr2 = s->usr2 & mask;
76
+ * Return an LED selector register value based on an existing one, with
67
77
+ * the appropriate 2-bit state value set for the given LED number (0-3).
68
- qemu_set_irq(s->irq, !!flags);
78
+ */
69
+ qemu_set_irq(s->irq, usr1 || usr2);
79
+static inline uint8_t pca955x_ledsel(uint8_t oldval, int led_num, int state)
80
+{
81
+ return (oldval & (~(0x3 << (led_num << 1)))) |
82
+ ((state & 0x3) << (led_num << 1));
83
+}
84
+
85
+static void pca9552_set_led(Object *obj, Visitor *v, const char *name,
86
+ void *opaque, Error **errp)
87
+{
88
+ PCA9552State *s = PCA9552(obj);
89
+ Error *local_err = NULL;
90
+ int led, rc, reg, val;
91
+ uint8_t state;
92
+ char *state_str;
93
+
94
+ visit_type_str(v, name, &state_str, &local_err);
95
+ if (local_err) {
96
+ error_propagate(errp, local_err);
97
+ return;
98
+ }
99
+ rc = sscanf(name, "led%2d", &led);
100
+ if (rc != 1) {
101
+ error_setg(errp, "%s: error reading %s", __func__, name);
102
+ return;
103
+ }
104
+ if (led < 0 || led > s->nr_leds) {
105
+ error_setg(errp, "%s invalid led %s", __func__, name);
106
+ return;
107
+ }
108
+
109
+ for (state = 0; state < ARRAY_SIZE(led_state); state++) {
110
+ if (!strcmp(state_str, led_state[state])) {
111
+ break;
112
+ }
113
+ }
114
+ if (state >= ARRAY_SIZE(led_state)) {
115
+ error_setg(errp, "%s invalid led state %s", __func__, state_str);
116
+ return;
117
+ }
118
+
119
+ reg = PCA9552_LS0 + led / 4;
120
+ val = pca9552_read(s, reg);
121
+ val = pca955x_ledsel(val, led % 4, state);
122
+ pca9552_write(s, reg, val);
123
+}
124
+
125
static const VMStateDescription pca9552_vmstate = {
126
.name = "PCA9552",
127
.version_id = 0,
128
@@ -XXX,XX +XXX,XX @@ static void pca9552_reset(DeviceState *dev)
129
static void pca9552_initfn(Object *obj)
130
{
131
PCA9552State *s = PCA9552(obj);
132
+ int led;
133
134
/* If support for the other PCA955X devices are implemented, these
135
* constant values might be part of class structure describing the
136
@@ -XXX,XX +XXX,XX @@ static void pca9552_initfn(Object *obj)
137
*/
138
s->max_reg = PCA9552_LS3;
139
s->nr_leds = 16;
140
+
141
+ for (led = 0; led < s->nr_leds; led++) {
142
+ char *name;
143
+
144
+ name = g_strdup_printf("led%d", led);
145
+ object_property_add(obj, name, "bool", pca9552_get_led, pca9552_set_led,
146
+ NULL, NULL, NULL);
147
+ g_free(name);
148
+ }
70
}
149
}
71
150
72
static void imx_serial_reset(IMXSerialState *s)
151
static void pca9552_class_init(ObjectClass *klass, void *data)
73
--
152
--
74
2.16.2
153
2.20.1
75
154
76
155
diff view generated by jsdifflib
1
Now we have separate types for BCM2386 and BCM2387, we might as well
1
From: Philippe Mathieu-Daudé <f4bug@amsat.org>
2
just hard-code the CPU type they use rather than having it passed
3
through as an object property. This then lets us put the initialization
4
of the CPU object in init rather than realize.
5
2
6
Note that this change means that it's no longer possible on
3
Since we enabled parallel TCG code generation for softmmu (see
7
the command line to use -cpu to ask for a different kind of
4
commit 3468b59 "tcg: enable multiple TCG contexts in softmmu")
8
CPU than the SoC supports. This was never a supported thing to
5
and its subsequent fix (commit 72649619 "add .min_cpus and
9
do anyway; we were just not sanity-checking the command line.
6
.default_cpus fields to machine_class"), the raspi machines are
7
restricted to always use their 4 cores:
10
8
11
This does require us to only build the bcm2837 object on
9
See in hw/arm/raspi2 (with BCM283X_NCPUS set to 4):
12
TARGET_AARCH64 configs, since otherwise it won't instantiate
13
due to the missing cortex-a53 device and "make check" will fail.
14
10
11
222 static void raspi2_machine_init(MachineClass *mc)
12
223 {
13
224 mc->desc = "Raspberry Pi 2";
14
230 mc->max_cpus = BCM283X_NCPUS;
15
231 mc->min_cpus = BCM283X_NCPUS;
16
232 mc->default_cpus = BCM283X_NCPUS;
17
235 };
18
236 DEFINE_MACHINE("raspi2", raspi2_machine_init)
19
20
We can no longer use the -smp option, as we get:
21
22
$ qemu-system-arm -M raspi2 -smp 1
23
qemu-system-arm: Invalid SMP CPUs 1. The min CPUs supported by machine 'raspi2' is 4
24
25
Since we can not set the TYPE_BCM283x SOC "enabled-cpus" with -smp,
26
remove the unuseful code.
27
28
We can achieve the same by using the '-global bcm2836.enabled-cpus=1'
29
option.
30
31
Reported-by: Laurent Bonnans <laurent.bonnans@here.com>
32
Signed-off-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
33
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
34
Message-id: 20200120235159.18510-2-f4bug@amsat.org
15
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
35
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
16
Reviewed-by: Andrew Baumann <Andrew.Baumann@microsoft.com>
17
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
18
Message-id: 20180313153458.26822-9-peter.maydell@linaro.org
19
---
36
---
20
hw/arm/bcm2836.c | 24 +++++++++++++++---------
37
hw/arm/raspi.c | 2 --
21
hw/arm/raspi.c | 2 --
38
1 file changed, 2 deletions(-)
22
2 files changed, 15 insertions(+), 11 deletions(-)
23
39
24
diff --git a/hw/arm/bcm2836.c b/hw/arm/bcm2836.c
25
index XXXXXXX..XXXXXXX 100644
26
--- a/hw/arm/bcm2836.c
27
+++ b/hw/arm/bcm2836.c
28
@@ -XXX,XX +XXX,XX @@
29
30
struct BCM283XInfo {
31
const char *name;
32
+ const char *cpu_type;
33
int clusterid;
34
};
35
36
static const BCM283XInfo bcm283x_socs[] = {
37
{
38
.name = TYPE_BCM2836,
39
+ .cpu_type = ARM_CPU_TYPE_NAME("cortex-a15"),
40
.clusterid = 0xf,
41
},
42
+#ifdef TARGET_AARCH64
43
{
44
.name = TYPE_BCM2837,
45
+ .cpu_type = ARM_CPU_TYPE_NAME("cortex-a53"),
46
.clusterid = 0x0,
47
},
48
+#endif
49
};
50
51
static void bcm2836_init(Object *obj)
52
{
53
BCM283XState *s = BCM283X(obj);
54
+ BCM283XClass *bc = BCM283X_GET_CLASS(obj);
55
+ const BCM283XInfo *info = bc->info;
56
+ int n;
57
+
58
+ for (n = 0; n < BCM283X_NCPUS; n++) {
59
+ object_initialize(&s->cpus[n], sizeof(s->cpus[n]),
60
+ info->cpu_type);
61
+ object_property_add_child(obj, "cpu[*]", OBJECT(&s->cpus[n]),
62
+ &error_abort);
63
+ }
64
65
object_initialize(&s->control, sizeof(s->control), TYPE_BCM2836_CONTROL);
66
object_property_add_child(obj, "control", OBJECT(&s->control), NULL);
67
@@ -XXX,XX +XXX,XX @@ static void bcm2836_realize(DeviceState *dev, Error **errp)
68
69
/* common peripherals from bcm2835 */
70
71
- obj = OBJECT(dev);
72
- for (n = 0; n < BCM283X_NCPUS; n++) {
73
- object_initialize(&s->cpus[n], sizeof(s->cpus[n]),
74
- s->cpu_type);
75
- object_property_add_child(obj, "cpu[*]", OBJECT(&s->cpus[n]),
76
- &error_abort);
77
- }
78
-
79
obj = object_property_get_link(OBJECT(dev), "ram", &err);
80
if (obj == NULL) {
81
error_setg(errp, "%s: required ram link not found: %s",
82
@@ -XXX,XX +XXX,XX @@ static void bcm2836_realize(DeviceState *dev, Error **errp)
83
}
84
85
static Property bcm2836_props[] = {
86
- DEFINE_PROP_STRING("cpu-type", BCM283XState, cpu_type),
87
DEFINE_PROP_UINT32("enabled-cpus", BCM283XState, enabled_cpus,
88
BCM283X_NCPUS),
89
DEFINE_PROP_END_OF_LIST()
90
diff --git a/hw/arm/raspi.c b/hw/arm/raspi.c
40
diff --git a/hw/arm/raspi.c b/hw/arm/raspi.c
91
index XXXXXXX..XXXXXXX 100644
41
index XXXXXXX..XXXXXXX 100644
92
--- a/hw/arm/raspi.c
42
--- a/hw/arm/raspi.c
93
+++ b/hw/arm/raspi.c
43
+++ b/hw/arm/raspi.c
94
@@ -XXX,XX +XXX,XX @@ static void raspi_init(MachineState *machine, int version)
44
@@ -XXX,XX +XXX,XX @@ static void raspi_init(MachineState *machine, int version)
95
/* Setup the SOC */
45
/* Setup the SOC */
96
object_property_add_const_link(OBJECT(&s->soc), "ram", OBJECT(&s->ram),
46
object_property_add_const_link(OBJECT(&s->soc), "ram", OBJECT(&s->ram),
97
&error_abort);
47
&error_abort);
98
- object_property_set_str(OBJECT(&s->soc), machine->cpu_type, "cpu-type",
48
- object_property_set_int(OBJECT(&s->soc), machine->smp.cpus, "enabled-cpus",
99
- &error_abort);
49
- &error_abort);
100
object_property_set_int(OBJECT(&s->soc), smp_cpus, "enabled-cpus",
50
int board_rev = version == 3 ? 0xa02082 : 0xa21041;
51
object_property_set_int(OBJECT(&s->soc), board_rev, "board-rev",
101
&error_abort);
52
&error_abort);
102
int board_rev = version == 3 ? 0xa02082 : 0xa21041;
103
--
53
--
104
2.16.2
54
2.20.1
105
55
106
56
diff view generated by jsdifflib
New patch
1
1
From: Damien Hedde <damien.hedde@greensocs.com>
2
3
Provide a temporary device_legacy_reset function doing what
4
device_reset does to prepare for the transition with Resettable
5
API.
6
7
All occurrence of device_reset in the code tree are also replaced
8
by device_legacy_reset.
9
10
The new resettable API has different prototype and semantics
11
(resetting child buses as well as the specified device). Subsequent
12
commits will make the changeover for each call site individually; once
13
that is complete device_legacy_reset() will be removed.
14
15
Signed-off-by: Damien Hedde <damien.hedde@greensocs.com>
16
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
17
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
18
Acked-by: David Gibson <david@gibson.dropbear.id.au>
19
Acked-by: Cornelia Huck <cohuck@redhat.com>
20
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
21
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
22
Message-id: 20200123132823.1117486-2-damien.hedde@greensocs.com
23
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
24
---
25
include/hw/qdev-core.h | 4 ++--
26
hw/audio/intel-hda.c | 2 +-
27
hw/core/qdev.c | 6 +++---
28
hw/hyperv/hyperv.c | 2 +-
29
hw/i386/microvm.c | 2 +-
30
hw/i386/pc.c | 2 +-
31
hw/ide/microdrive.c | 8 ++++----
32
hw/intc/spapr_xive.c | 2 +-
33
hw/ppc/pnv_psi.c | 4 ++--
34
hw/ppc/spapr_pci.c | 2 +-
35
hw/ppc/spapr_vio.c | 2 +-
36
hw/s390x/s390-pci-inst.c | 2 +-
37
hw/scsi/vmw_pvscsi.c | 2 +-
38
hw/sd/omap_mmc.c | 2 +-
39
hw/sd/pl181.c | 2 +-
40
15 files changed, 22 insertions(+), 22 deletions(-)
41
42
diff --git a/include/hw/qdev-core.h b/include/hw/qdev-core.h
43
index XXXXXXX..XXXXXXX 100644
44
--- a/include/hw/qdev-core.h
45
+++ b/include/hw/qdev-core.h
46
@@ -XXX,XX +XXX,XX @@ char *qdev_get_own_fw_dev_path_from_handler(BusState *bus, DeviceState *dev);
47
void qdev_machine_init(void);
48
49
/**
50
- * @device_reset
51
+ * device_legacy_reset:
52
*
53
* Reset a single device (by calling the reset method).
54
*/
55
-void device_reset(DeviceState *dev);
56
+void device_legacy_reset(DeviceState *dev);
57
58
void device_class_set_props(DeviceClass *dc, Property *props);
59
60
diff --git a/hw/audio/intel-hda.c b/hw/audio/intel-hda.c
61
index XXXXXXX..XXXXXXX 100644
62
--- a/hw/audio/intel-hda.c
63
+++ b/hw/audio/intel-hda.c
64
@@ -XXX,XX +XXX,XX @@ static void intel_hda_reset(DeviceState *dev)
65
QTAILQ_FOREACH(kid, &d->codecs.qbus.children, sibling) {
66
DeviceState *qdev = kid->child;
67
cdev = HDA_CODEC_DEVICE(qdev);
68
- device_reset(DEVICE(cdev));
69
+ device_legacy_reset(DEVICE(cdev));
70
d->state_sts |= (1 << cdev->cad);
71
}
72
intel_hda_update_irq(d);
73
diff --git a/hw/core/qdev.c b/hw/core/qdev.c
74
index XXXXXXX..XXXXXXX 100644
75
--- a/hw/core/qdev.c
76
+++ b/hw/core/qdev.c
77
@@ -XXX,XX +XXX,XX @@ HotplugHandler *qdev_get_hotplug_handler(DeviceState *dev)
78
79
static int qdev_reset_one(DeviceState *dev, void *opaque)
80
{
81
- device_reset(dev);
82
+ device_legacy_reset(dev);
83
84
return 0;
85
}
86
@@ -XXX,XX +XXX,XX @@ static void device_set_realized(Object *obj, bool value, Error **errp)
87
}
88
}
89
if (dev->hotplugged) {
90
- device_reset(dev);
91
+ device_legacy_reset(dev);
92
}
93
dev->pending_deleted_event = false;
94
95
@@ -XXX,XX +XXX,XX @@ void device_class_set_parent_unrealize(DeviceClass *dc,
96
dc->unrealize = dev_unrealize;
97
}
98
99
-void device_reset(DeviceState *dev)
100
+void device_legacy_reset(DeviceState *dev)
101
{
102
DeviceClass *klass = DEVICE_GET_CLASS(dev);
103
104
diff --git a/hw/hyperv/hyperv.c b/hw/hyperv/hyperv.c
105
index XXXXXXX..XXXXXXX 100644
106
--- a/hw/hyperv/hyperv.c
107
+++ b/hw/hyperv/hyperv.c
108
@@ -XXX,XX +XXX,XX @@ void hyperv_synic_reset(CPUState *cs)
109
SynICState *synic = get_synic(cs);
110
111
if (synic) {
112
- device_reset(DEVICE(synic));
113
+ device_legacy_reset(DEVICE(synic));
114
}
115
}
116
117
diff --git a/hw/i386/microvm.c b/hw/i386/microvm.c
118
index XXXXXXX..XXXXXXX 100644
119
--- a/hw/i386/microvm.c
120
+++ b/hw/i386/microvm.c
121
@@ -XXX,XX +XXX,XX @@ static void microvm_machine_reset(MachineState *machine)
122
cpu = X86_CPU(cs);
123
124
if (cpu->apic_state) {
125
- device_reset(cpu->apic_state);
126
+ device_legacy_reset(cpu->apic_state);
127
}
128
}
129
}
130
diff --git a/hw/i386/pc.c b/hw/i386/pc.c
131
index XXXXXXX..XXXXXXX 100644
132
--- a/hw/i386/pc.c
133
+++ b/hw/i386/pc.c
134
@@ -XXX,XX +XXX,XX @@ static void pc_machine_reset(MachineState *machine)
135
cpu = X86_CPU(cs);
136
137
if (cpu->apic_state) {
138
- device_reset(cpu->apic_state);
139
+ device_legacy_reset(cpu->apic_state);
140
}
141
}
142
}
143
diff --git a/hw/ide/microdrive.c b/hw/ide/microdrive.c
144
index XXXXXXX..XXXXXXX 100644
145
--- a/hw/ide/microdrive.c
146
+++ b/hw/ide/microdrive.c
147
@@ -XXX,XX +XXX,XX @@ static void md_attr_write(PCMCIACardState *card, uint32_t at, uint8_t value)
148
case 0x00:    /* Configuration Option Register */
149
s->opt = value & 0xcf;
150
if (value & OPT_SRESET) {
151
- device_reset(DEVICE(s));
152
+ device_legacy_reset(DEVICE(s));
153
}
154
md_interrupt_update(s);
155
break;
156
@@ -XXX,XX +XXX,XX @@ static void md_common_write(PCMCIACardState *card, uint32_t at, uint16_t value)
157
case 0xe:    /* Device Control */
158
s->ctrl = value;
159
if (value & CTRL_SRST) {
160
- device_reset(DEVICE(s));
161
+ device_legacy_reset(DEVICE(s));
162
}
163
md_interrupt_update(s);
164
break;
165
@@ -XXX,XX +XXX,XX @@ static int dscm1xxxx_attach(PCMCIACardState *card)
166
md->attr_base = pcc->cis[0x74] | (pcc->cis[0x76] << 8);
167
md->io_base = 0x0;
168
169
- device_reset(DEVICE(md));
170
+ device_legacy_reset(DEVICE(md));
171
md_interrupt_update(md);
172
173
return 0;
174
@@ -XXX,XX +XXX,XX @@ static int dscm1xxxx_detach(PCMCIACardState *card)
175
{
176
MicroDriveState *md = MICRODRIVE(card);
177
178
- device_reset(DEVICE(md));
179
+ device_legacy_reset(DEVICE(md));
180
return 0;
181
}
182
183
diff --git a/hw/intc/spapr_xive.c b/hw/intc/spapr_xive.c
184
index XXXXXXX..XXXXXXX 100644
185
--- a/hw/intc/spapr_xive.c
186
+++ b/hw/intc/spapr_xive.c
187
@@ -XXX,XX +XXX,XX @@ static target_ulong h_int_reset(PowerPCCPU *cpu,
188
return H_PARAMETER;
189
}
190
191
- device_reset(DEVICE(xive));
192
+ device_legacy_reset(DEVICE(xive));
193
194
if (kvm_irqchip_in_kernel()) {
195
Error *local_err = NULL;
196
diff --git a/hw/ppc/pnv_psi.c b/hw/ppc/pnv_psi.c
197
index XXXXXXX..XXXXXXX 100644
198
--- a/hw/ppc/pnv_psi.c
199
+++ b/hw/ppc/pnv_psi.c
200
@@ -XXX,XX +XXX,XX @@ static void pnv_psi_reset(DeviceState *dev)
201
202
static void pnv_psi_reset_handler(void *dev)
203
{
204
- device_reset(DEVICE(dev));
205
+ device_legacy_reset(DEVICE(dev));
206
}
207
208
static void pnv_psi_realize(DeviceState *dev, Error **errp)
209
@@ -XXX,XX +XXX,XX @@ static void pnv_psi_p9_mmio_write(void *opaque, hwaddr addr,
210
break;
211
case PSIHB9_INTERRUPT_CONTROL:
212
if (val & PSIHB9_IRQ_RESET) {
213
- device_reset(DEVICE(&psi9->source));
214
+ device_legacy_reset(DEVICE(&psi9->source));
215
}
216
psi->regs[reg] = val;
217
break;
218
diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
219
index XXXXXXX..XXXXXXX 100644
220
--- a/hw/ppc/spapr_pci.c
221
+++ b/hw/ppc/spapr_pci.c
222
@@ -XXX,XX +XXX,XX @@ static int spapr_phb_children_reset(Object *child, void *opaque)
223
DeviceState *dev = (DeviceState *) object_dynamic_cast(child, TYPE_DEVICE);
224
225
if (dev) {
226
- device_reset(dev);
227
+ device_legacy_reset(dev);
228
}
229
230
return 0;
231
diff --git a/hw/ppc/spapr_vio.c b/hw/ppc/spapr_vio.c
232
index XXXXXXX..XXXXXXX 100644
233
--- a/hw/ppc/spapr_vio.c
234
+++ b/hw/ppc/spapr_vio.c
235
@@ -XXX,XX +XXX,XX @@ int spapr_vio_send_crq(SpaprVioDevice *dev, uint8_t *crq)
236
static void spapr_vio_quiesce_one(SpaprVioDevice *dev)
237
{
238
if (dev->tcet) {
239
- device_reset(DEVICE(dev->tcet));
240
+ device_legacy_reset(DEVICE(dev->tcet));
241
}
242
free_crq(dev);
243
}
244
diff --git a/hw/s390x/s390-pci-inst.c b/hw/s390x/s390-pci-inst.c
245
index XXXXXXX..XXXXXXX 100644
246
--- a/hw/s390x/s390-pci-inst.c
247
+++ b/hw/s390x/s390-pci-inst.c
248
@@ -XXX,XX +XXX,XX @@ int clp_service_call(S390CPU *cpu, uint8_t r2, uintptr_t ra)
249
stw_p(&ressetpci->hdr.rsp, CLP_RC_SETPCIFN_FHOP);
250
goto out;
251
}
252
- device_reset(DEVICE(pbdev));
253
+ device_legacy_reset(DEVICE(pbdev));
254
pbdev->fh &= ~FH_MASK_ENABLE;
255
pbdev->state = ZPCI_FS_DISABLED;
256
stl_p(&ressetpci->fh, pbdev->fh);
257
diff --git a/hw/scsi/vmw_pvscsi.c b/hw/scsi/vmw_pvscsi.c
258
index XXXXXXX..XXXXXXX 100644
259
--- a/hw/scsi/vmw_pvscsi.c
260
+++ b/hw/scsi/vmw_pvscsi.c
261
@@ -XXX,XX +XXX,XX @@ pvscsi_on_cmd_reset_device(PVSCSIState *s)
262
263
if (sdev != NULL) {
264
s->resetting++;
265
- device_reset(&sdev->qdev);
266
+ device_legacy_reset(&sdev->qdev);
267
s->resetting--;
268
return PVSCSI_COMMAND_PROCESSING_SUCCEEDED;
269
}
270
diff --git a/hw/sd/omap_mmc.c b/hw/sd/omap_mmc.c
271
index XXXXXXX..XXXXXXX 100644
272
--- a/hw/sd/omap_mmc.c
273
+++ b/hw/sd/omap_mmc.c
274
@@ -XXX,XX +XXX,XX @@ void omap_mmc_reset(struct omap_mmc_s *host)
275
* into any bus, and we must reset it manually. When omap_mmc is
276
* QOMified this must move into the QOM reset function.
277
*/
278
- device_reset(DEVICE(host->card));
279
+ device_legacy_reset(DEVICE(host->card));
280
}
281
282
static uint64_t omap_mmc_read(void *opaque, hwaddr offset,
283
diff --git a/hw/sd/pl181.c b/hw/sd/pl181.c
284
index XXXXXXX..XXXXXXX 100644
285
--- a/hw/sd/pl181.c
286
+++ b/hw/sd/pl181.c
287
@@ -XXX,XX +XXX,XX @@ static void pl181_reset(DeviceState *d)
288
/* Since we're still using the legacy SD API the card is not plugged
289
* into any bus, and we must reset it manually.
290
*/
291
- device_reset(DEVICE(s->card));
292
+ device_legacy_reset(DEVICE(s->card));
293
}
294
295
static void pl181_init(Object *obj)
296
--
297
2.20.1
298
299
diff view generated by jsdifflib
New patch
1
From: Damien Hedde <damien.hedde@greensocs.com>
1
2
3
Adds trace events to reset procedure and when updating the parent
4
bus of a device.
5
6
Signed-off-by: Damien Hedde <damien.hedde@greensocs.com>
7
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
8
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
9
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
10
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
11
Message-id: 20200123132823.1117486-3-damien.hedde@greensocs.com
12
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
13
---
14
hw/core/qdev.c | 29 ++++++++++++++++++++++++++---
15
hw/core/trace-events | 9 +++++++++
16
2 files changed, 35 insertions(+), 3 deletions(-)
17
18
diff --git a/hw/core/qdev.c b/hw/core/qdev.c
19
index XXXXXXX..XXXXXXX 100644
20
--- a/hw/core/qdev.c
21
+++ b/hw/core/qdev.c
22
@@ -XXX,XX +XXX,XX @@
23
#include "hw/boards.h"
24
#include "hw/sysbus.h"
25
#include "migration/vmstate.h"
26
+#include "trace.h"
27
28
bool qdev_hotplug = false;
29
static bool qdev_hot_added = false;
30
@@ -XXX,XX +XXX,XX @@ void qdev_set_parent_bus(DeviceState *dev, BusState *bus)
31
bool replugging = dev->parent_bus != NULL;
32
33
if (replugging) {
34
- /* Keep a reference to the device while it's not plugged into
35
+ trace_qdev_update_parent_bus(dev, object_get_typename(OBJECT(dev)),
36
+ dev->parent_bus, object_get_typename(OBJECT(dev->parent_bus)),
37
+ OBJECT(bus), object_get_typename(OBJECT(bus)));
38
+ /*
39
+ * Keep a reference to the device while it's not plugged into
40
* any bus, to avoid it potentially evaporating when it is
41
* dereffed in bus_remove_child().
42
*/
43
@@ -XXX,XX +XXX,XX @@ HotplugHandler *qdev_get_hotplug_handler(DeviceState *dev)
44
return hotplug_ctrl;
45
}
46
47
+static int qdev_prereset(DeviceState *dev, void *opaque)
48
+{
49
+ trace_qdev_reset_tree(dev, object_get_typename(OBJECT(dev)));
50
+ return 0;
51
+}
52
+
53
+static int qbus_prereset(BusState *bus, void *opaque)
54
+{
55
+ trace_qbus_reset_tree(bus, object_get_typename(OBJECT(bus)));
56
+ return 0;
57
+}
58
+
59
static int qdev_reset_one(DeviceState *dev, void *opaque)
60
{
61
device_legacy_reset(dev);
62
@@ -XXX,XX +XXX,XX @@ static int qdev_reset_one(DeviceState *dev, void *opaque)
63
static int qbus_reset_one(BusState *bus, void *opaque)
64
{
65
BusClass *bc = BUS_GET_CLASS(bus);
66
+ trace_qbus_reset(bus, object_get_typename(OBJECT(bus)));
67
if (bc->reset) {
68
bc->reset(bus);
69
}
70
@@ -XXX,XX +XXX,XX @@ static int qbus_reset_one(BusState *bus, void *opaque)
71
72
void qdev_reset_all(DeviceState *dev)
73
{
74
- qdev_walk_children(dev, NULL, NULL, qdev_reset_one, qbus_reset_one, NULL);
75
+ trace_qdev_reset_all(dev, object_get_typename(OBJECT(dev)));
76
+ qdev_walk_children(dev, qdev_prereset, qbus_prereset,
77
+ qdev_reset_one, qbus_reset_one, NULL);
78
}
79
80
void qdev_reset_all_fn(void *opaque)
81
@@ -XXX,XX +XXX,XX @@ void qdev_reset_all_fn(void *opaque)
82
83
void qbus_reset_all(BusState *bus)
84
{
85
- qbus_walk_children(bus, NULL, NULL, qdev_reset_one, qbus_reset_one, NULL);
86
+ trace_qbus_reset_all(bus, object_get_typename(OBJECT(bus)));
87
+ qbus_walk_children(bus, qdev_prereset, qbus_prereset,
88
+ qdev_reset_one, qbus_reset_one, NULL);
89
}
90
91
void qbus_reset_all_fn(void *opaque)
92
@@ -XXX,XX +XXX,XX @@ void device_legacy_reset(DeviceState *dev)
93
{
94
DeviceClass *klass = DEVICE_GET_CLASS(dev);
95
96
+ trace_qdev_reset(dev, object_get_typename(OBJECT(dev)));
97
if (klass->reset) {
98
klass->reset(dev);
99
}
100
diff --git a/hw/core/trace-events b/hw/core/trace-events
101
index XXXXXXX..XXXXXXX 100644
102
--- a/hw/core/trace-events
103
+++ b/hw/core/trace-events
104
@@ -XXX,XX +XXX,XX @@
105
# loader.c
106
loader_write_rom(const char *name, uint64_t gpa, uint64_t size, bool isrom) "%s: @0x%"PRIx64" size=0x%"PRIx64" ROM=%d"
107
+
108
+# qdev.c
109
+qdev_reset(void *obj, const char *objtype) "obj=%p(%s)"
110
+qdev_reset_all(void *obj, const char *objtype) "obj=%p(%s)"
111
+qdev_reset_tree(void *obj, const char *objtype) "obj=%p(%s)"
112
+qbus_reset(void *obj, const char *objtype) "obj=%p(%s)"
113
+qbus_reset_all(void *obj, const char *objtype) "obj=%p(%s)"
114
+qbus_reset_tree(void *obj, const char *objtype) "obj=%p(%s)"
115
+qdev_update_parent_bus(void *obj, const char *objtype, void *oldp, const char *oldptype, void *newp, const char *newptype) "obj=%p(%s) old_parent=%p(%s) new_parent=%p(%s)"
116
--
117
2.20.1
118
119
diff view generated by jsdifflib
New patch
1
1
From: Damien Hedde <damien.hedde@greensocs.com>
2
3
This commit defines an interface allowing multi-phase reset. This aims
4
to solve a problem of the actual single-phase reset (built in
5
DeviceClass and BusClass): reset behavior is dependent on the order
6
in which reset handlers are called. In particular doing external
7
side-effect (like setting an qemu_irq) is problematic because receiving
8
object may not be reset yet.
9
10
The Resettable interface divides the reset in 3 well defined phases.
11
To reset an object tree, all 1st phases are executed then all 2nd then
12
all 3rd. See the comments in include/hw/resettable.h for a more complete
13
description. The interface defines 3 phases to let the future
14
possibility of holding an object into reset for some time.
15
16
The qdev/qbus reset in DeviceClass and BusClass will be modified in
17
following commits to use this interface. A mechanism is provided
18
to allow executing a transitional reset handler in place of the 2nd
19
phase which is executed in children-then-parent order inside a tree.
20
This will allow to transition devices and buses smoothly while
21
keeping the exact current qdev/qbus reset behavior for now.
22
23
Documentation will be added in a following commit.
24
25
Signed-off-by: Damien Hedde <damien.hedde@greensocs.com>
26
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
27
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
28
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
29
Message-id: 20200123132823.1117486-4-damien.hedde@greensocs.com
30
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
31
---
32
hw/core/Makefile.objs | 1 +
33
include/hw/resettable.h | 211 +++++++++++++++++++++++++++++++++++
34
hw/core/resettable.c | 238 ++++++++++++++++++++++++++++++++++++++++
35
hw/core/trace-events | 17 +++
36
4 files changed, 467 insertions(+)
37
create mode 100644 include/hw/resettable.h
38
create mode 100644 hw/core/resettable.c
39
40
diff --git a/hw/core/Makefile.objs b/hw/core/Makefile.objs
41
index XXXXXXX..XXXXXXX 100644
42
--- a/hw/core/Makefile.objs
43
+++ b/hw/core/Makefile.objs
44
@@ -XXX,XX +XXX,XX @@
45
common-obj-y += qdev.o qdev-properties.o
46
common-obj-y += bus.o
47
common-obj-y += cpu.o
48
+common-obj-y += resettable.o
49
common-obj-y += hotplug.o
50
common-obj-y += vmstate-if.o
51
# irq.o needed for qdev GPIO handling:
52
diff --git a/include/hw/resettable.h b/include/hw/resettable.h
53
new file mode 100644
54
index XXXXXXX..XXXXXXX
55
--- /dev/null
56
+++ b/include/hw/resettable.h
57
@@ -XXX,XX +XXX,XX @@
58
+/*
59
+ * Resettable interface header.
60
+ *
61
+ * Copyright (c) 2019 GreenSocs SAS
62
+ *
63
+ * Authors:
64
+ * Damien Hedde
65
+ *
66
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
67
+ * See the COPYING file in the top-level directory.
68
+ */
69
+
70
+#ifndef HW_RESETTABLE_H
71
+#define HW_RESETTABLE_H
72
+
73
+#include "qom/object.h"
74
+
75
+#define TYPE_RESETTABLE_INTERFACE "resettable"
76
+
77
+#define RESETTABLE_CLASS(class) \
78
+ OBJECT_CLASS_CHECK(ResettableClass, (class), TYPE_RESETTABLE_INTERFACE)
79
+
80
+#define RESETTABLE_GET_CLASS(obj) \
81
+ OBJECT_GET_CLASS(ResettableClass, (obj), TYPE_RESETTABLE_INTERFACE)
82
+
83
+typedef struct ResettableState ResettableState;
84
+
85
+/**
86
+ * ResetType:
87
+ * Types of reset.
88
+ *
89
+ * + Cold: reset resulting from a power cycle of the object.
90
+ *
91
+ * TODO: Support has to be added to handle more types. In particular,
92
+ * ResettableState structure needs to be expanded.
93
+ */
94
+typedef enum ResetType {
95
+ RESET_TYPE_COLD,
96
+} ResetType;
97
+
98
+/*
99
+ * ResettableClass:
100
+ * Interface for resettable objects.
101
+ *
102
+ * See docs/devel/reset.rst for more detailed information about how QEMU models
103
+ * reset. This whole API must only be used when holding the iothread mutex.
104
+ *
105
+ * All objects which can be reset must implement this interface;
106
+ * it is usually provided by a base class such as DeviceClass or BusClass.
107
+ * Every Resettable object must maintain some state tracking the
108
+ * progress of a reset operation by providing a ResettableState structure.
109
+ * The functions defined in this module take care of updating the
110
+ * state of the reset.
111
+ * The base class implementation of the interface provides this
112
+ * state and implements the associated method: get_state.
113
+ *
114
+ * Concrete object implementations (typically specific devices
115
+ * such as a UART model) should provide the functions
116
+ * for the phases.enter, phases.hold and phases.exit methods, which
117
+ * they can set in their class init function, either directly or
118
+ * by calling resettable_class_set_parent_phases().
119
+ * The phase methods are guaranteed to only only ever be called once
120
+ * for any reset event, in the order 'enter', 'hold', 'exit'.
121
+ * An object will always move quickly from 'enter' to 'hold'
122
+ * but might remain in 'hold' for an arbitrary period of time
123
+ * before eventually reset is deasserted and the 'exit' phase is called.
124
+ * Object implementations should be prepared for functions handling
125
+ * inbound connections from other devices (such as qemu_irq handler
126
+ * functions) to be called at any point during reset after their
127
+ * 'enter' method has been called.
128
+ *
129
+ * Users of a resettable object should not call these methods
130
+ * directly, but instead use the function resettable_reset().
131
+ *
132
+ * @phases.enter: This phase is called when the object enters reset. It
133
+ * should reset local state of the object, but it must not do anything that
134
+ * has a side-effect on other objects, such as raising or lowering a qemu_irq
135
+ * line or reading or writing guest memory. It takes the reset's type as
136
+ * argument.
137
+ *
138
+ * @phases.hold: This phase is called for entry into reset, once every object
139
+ * in the system which is being reset has had its @phases.enter method called.
140
+ * At this point devices can do actions that affect other objects.
141
+ *
142
+ * @phases.exit: This phase is called when the object leaves the reset state.
143
+ * Actions affecting other objects are permitted.
144
+ *
145
+ * @get_state: Mandatory method which must return a pointer to a
146
+ * ResettableState.
147
+ *
148
+ * @get_transitional_function: transitional method to handle Resettable objects
149
+ * not yet fully moved to this interface. It will be removed as soon as it is
150
+ * not needed anymore. This method is optional and may return a pointer to a
151
+ * function to be used instead of the phases. If the method exists and returns
152
+ * a non-NULL function pointer then that function is executed as a replacement
153
+ * of the 'hold' phase method taking the object as argument. The two other phase
154
+ * methods are not executed.
155
+ *
156
+ * @child_foreach: Executes a given callback on every Resettable child. Child
157
+ * in this context means a child in the qbus tree, so the children of a qbus
158
+ * are the devices on it, and the children of a device are all the buses it
159
+ * owns. This is not the same as the QOM object hierarchy. The function takes
160
+ * additional opaque and ResetType arguments which must be passed unmodified to
161
+ * the callback.
162
+ */
163
+typedef void (*ResettableEnterPhase)(Object *obj, ResetType type);
164
+typedef void (*ResettableHoldPhase)(Object *obj);
165
+typedef void (*ResettableExitPhase)(Object *obj);
166
+typedef ResettableState * (*ResettableGetState)(Object *obj);
167
+typedef void (*ResettableTrFunction)(Object *obj);
168
+typedef ResettableTrFunction (*ResettableGetTrFunction)(Object *obj);
169
+typedef void (*ResettableChildCallback)(Object *, void *opaque,
170
+ ResetType type);
171
+typedef void (*ResettableChildForeach)(Object *obj,
172
+ ResettableChildCallback cb,
173
+ void *opaque, ResetType type);
174
+typedef struct ResettablePhases {
175
+ ResettableEnterPhase enter;
176
+ ResettableHoldPhase hold;
177
+ ResettableExitPhase exit;
178
+} ResettablePhases;
179
+typedef struct ResettableClass {
180
+ InterfaceClass parent_class;
181
+
182
+ /* Phase methods */
183
+ ResettablePhases phases;
184
+
185
+ /* State access method */
186
+ ResettableGetState get_state;
187
+
188
+ /* Transitional method for legacy reset compatibility */
189
+ ResettableGetTrFunction get_transitional_function;
190
+
191
+ /* Hierarchy handling method */
192
+ ResettableChildForeach child_foreach;
193
+} ResettableClass;
194
+
195
+/**
196
+ * ResettableState:
197
+ * Structure holding reset related state. The fields should not be accessed
198
+ * directly; the definition is here to allow further inclusion into other
199
+ * objects.
200
+ *
201
+ * @count: Number of reset level the object is into. It is incremented when
202
+ * the reset operation starts and decremented when it finishes.
203
+ * @hold_phase_pending: flag which indicates that we need to invoke the 'hold'
204
+ * phase handler for this object.
205
+ * @exit_phase_in_progress: true if we are currently in the exit phase
206
+ */
207
+struct ResettableState {
208
+ unsigned count;
209
+ bool hold_phase_pending;
210
+ bool exit_phase_in_progress;
211
+};
212
+
213
+/**
214
+ * resettable_reset:
215
+ * Trigger a reset on an object @obj of type @type. @obj must implement
216
+ * Resettable interface.
217
+ *
218
+ * Calling this function is equivalent to calling @resettable_assert_reset()
219
+ * then @resettable_release_reset().
220
+ */
221
+void resettable_reset(Object *obj, ResetType type);
222
+
223
+/**
224
+ * resettable_assert_reset:
225
+ * Put an object @obj into reset. @obj must implement Resettable interface.
226
+ *
227
+ * @resettable_release_reset() must eventually be called after this call.
228
+ * There must be one call to @resettable_release_reset() per call of
229
+ * @resettable_assert_reset(), with the same type argument.
230
+ *
231
+ * NOTE: Until support for migration is added, the @resettable_release_reset()
232
+ * must not be delayed. It must occur just after @resettable_assert_reset() so
233
+ * that migration cannot be triggered in between. Prefer using
234
+ * @resettable_reset() for now.
235
+ */
236
+void resettable_assert_reset(Object *obj, ResetType type);
237
+
238
+/**
239
+ * resettable_release_reset:
240
+ * Release the object @obj from reset. @obj must implement Resettable interface.
241
+ *
242
+ * See @resettable_assert_reset() description for details.
243
+ */
244
+void resettable_release_reset(Object *obj, ResetType type);
245
+
246
+/**
247
+ * resettable_is_in_reset:
248
+ * Return true if @obj is under reset.
249
+ *
250
+ * @obj must implement Resettable interface.
251
+ */
252
+bool resettable_is_in_reset(Object *obj);
253
+
254
+/**
255
+ * resettable_class_set_parent_phases:
256
+ *
257
+ * Save @rc current reset phases into @parent_phases and override @rc phases
258
+ * by the given new methods (@enter, @hold and @exit).
259
+ * Each phase is overridden only if the new one is not NULL allowing to
260
+ * override a subset of phases.
261
+ */
262
+void resettable_class_set_parent_phases(ResettableClass *rc,
263
+ ResettableEnterPhase enter,
264
+ ResettableHoldPhase hold,
265
+ ResettableExitPhase exit,
266
+ ResettablePhases *parent_phases);
267
+
268
+#endif
269
diff --git a/hw/core/resettable.c b/hw/core/resettable.c
270
new file mode 100644
271
index XXXXXXX..XXXXXXX
272
--- /dev/null
273
+++ b/hw/core/resettable.c
274
@@ -XXX,XX +XXX,XX @@
275
+/*
276
+ * Resettable interface.
277
+ *
278
+ * Copyright (c) 2019 GreenSocs SAS
279
+ *
280
+ * Authors:
281
+ * Damien Hedde
282
+ *
283
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
284
+ * See the COPYING file in the top-level directory.
285
+ */
286
+
287
+#include "qemu/osdep.h"
288
+#include "qemu/module.h"
289
+#include "hw/resettable.h"
290
+#include "trace.h"
291
+
292
+/**
293
+ * resettable_phase_enter/hold/exit:
294
+ * Function executing a phase recursively in a resettable object and its
295
+ * children.
296
+ */
297
+static void resettable_phase_enter(Object *obj, void *opaque, ResetType type);
298
+static void resettable_phase_hold(Object *obj, void *opaque, ResetType type);
299
+static void resettable_phase_exit(Object *obj, void *opaque, ResetType type);
300
+
301
+/**
302
+ * enter_phase_in_progress:
303
+ * True if we are currently in reset enter phase.
304
+ *
305
+ * Note: This flag is only used to guarantee (using asserts) that the reset
306
+ * API is used correctly. We can use a global variable because we rely on the
307
+ * iothread mutex to ensure only one reset operation is in a progress at a
308
+ * given time.
309
+ */
310
+static bool enter_phase_in_progress;
311
+
312
+void resettable_reset(Object *obj, ResetType type)
313
+{
314
+ trace_resettable_reset(obj, type);
315
+ resettable_assert_reset(obj, type);
316
+ resettable_release_reset(obj, type);
317
+}
318
+
319
+void resettable_assert_reset(Object *obj, ResetType type)
320
+{
321
+ /* TODO: change this assert when adding support for other reset types */
322
+ assert(type == RESET_TYPE_COLD);
323
+ trace_resettable_reset_assert_begin(obj, type);
324
+ assert(!enter_phase_in_progress);
325
+
326
+ enter_phase_in_progress = true;
327
+ resettable_phase_enter(obj, NULL, type);
328
+ enter_phase_in_progress = false;
329
+
330
+ resettable_phase_hold(obj, NULL, type);
331
+
332
+ trace_resettable_reset_assert_end(obj);
333
+}
334
+
335
+void resettable_release_reset(Object *obj, ResetType type)
336
+{
337
+ /* TODO: change this assert when adding support for other reset types */
338
+ assert(type == RESET_TYPE_COLD);
339
+ trace_resettable_reset_release_begin(obj, type);
340
+ assert(!enter_phase_in_progress);
341
+
342
+ resettable_phase_exit(obj, NULL, type);
343
+
344
+ trace_resettable_reset_release_end(obj);
345
+}
346
+
347
+bool resettable_is_in_reset(Object *obj)
348
+{
349
+ ResettableClass *rc = RESETTABLE_GET_CLASS(obj);
350
+ ResettableState *s = rc->get_state(obj);
351
+
352
+ return s->count > 0;
353
+}
354
+
355
+/**
356
+ * resettable_child_foreach:
357
+ * helper to avoid checking the existence of the method.
358
+ */
359
+static void resettable_child_foreach(ResettableClass *rc, Object *obj,
360
+ ResettableChildCallback cb,
361
+ void *opaque, ResetType type)
362
+{
363
+ if (rc->child_foreach) {
364
+ rc->child_foreach(obj, cb, opaque, type);
365
+ }
366
+}
367
+
368
+/**
369
+ * resettable_get_tr_func:
370
+ * helper to fetch transitional reset callback if any.
371
+ */
372
+static ResettableTrFunction resettable_get_tr_func(ResettableClass *rc,
373
+ Object *obj)
374
+{
375
+ ResettableTrFunction tr_func = NULL;
376
+ if (rc->get_transitional_function) {
377
+ tr_func = rc->get_transitional_function(obj);
378
+ }
379
+ return tr_func;
380
+}
381
+
382
+static void resettable_phase_enter(Object *obj, void *opaque, ResetType type)
383
+{
384
+ ResettableClass *rc = RESETTABLE_GET_CLASS(obj);
385
+ ResettableState *s = rc->get_state(obj);
386
+ const char *obj_typename = object_get_typename(obj);
387
+ bool action_needed = false;
388
+
389
+ /* exit phase has to finish properly before entering back in reset */
390
+ assert(!s->exit_phase_in_progress);
391
+
392
+ trace_resettable_phase_enter_begin(obj, obj_typename, s->count, type);
393
+
394
+ /* Only take action if we really enter reset for the 1st time. */
395
+ /*
396
+ * TODO: if adding more ResetType support, some additional checks
397
+ * are probably needed here.
398
+ */
399
+ if (s->count++ == 0) {
400
+ action_needed = true;
401
+ }
402
+ /*
403
+ * We limit the count to an arbitrary "big" value. The value is big
404
+ * enough not to be triggered normally.
405
+ * The assert will stop an infinite loop if there is a cycle in the
406
+ * reset tree. The loop goes through resettable_foreach_child below
407
+ * which at some point will call us again.
408
+ */
409
+ assert(s->count <= 50);
410
+
411
+ /*
412
+ * handle the children even if action_needed is at false so that
413
+ * child counts are incremented too
414
+ */
415
+ resettable_child_foreach(rc, obj, resettable_phase_enter, NULL, type);
416
+
417
+ /* execute enter phase for the object if needed */
418
+ if (action_needed) {
419
+ trace_resettable_phase_enter_exec(obj, obj_typename, type,
420
+ !!rc->phases.enter);
421
+ if (rc->phases.enter && !resettable_get_tr_func(rc, obj)) {
422
+ rc->phases.enter(obj, type);
423
+ }
424
+ s->hold_phase_pending = true;
425
+ }
426
+ trace_resettable_phase_enter_end(obj, obj_typename, s->count);
427
+}
428
+
429
+static void resettable_phase_hold(Object *obj, void *opaque, ResetType type)
430
+{
431
+ ResettableClass *rc = RESETTABLE_GET_CLASS(obj);
432
+ ResettableState *s = rc->get_state(obj);
433
+ const char *obj_typename = object_get_typename(obj);
434
+
435
+ /* exit phase has to finish properly before entering back in reset */
436
+ assert(!s->exit_phase_in_progress);
437
+
438
+ trace_resettable_phase_hold_begin(obj, obj_typename, s->count, type);
439
+
440
+ /* handle children first */
441
+ resettable_child_foreach(rc, obj, resettable_phase_hold, NULL, type);
442
+
443
+ /* exec hold phase */
444
+ if (s->hold_phase_pending) {
445
+ s->hold_phase_pending = false;
446
+ ResettableTrFunction tr_func = resettable_get_tr_func(rc, obj);
447
+ trace_resettable_phase_hold_exec(obj, obj_typename, !!rc->phases.hold);
448
+ if (tr_func) {
449
+ trace_resettable_transitional_function(obj, obj_typename);
450
+ tr_func(obj);
451
+ } else if (rc->phases.hold) {
452
+ rc->phases.hold(obj);
453
+ }
454
+ }
455
+ trace_resettable_phase_hold_end(obj, obj_typename, s->count);
456
+}
457
+
458
+static void resettable_phase_exit(Object *obj, void *opaque, ResetType type)
459
+{
460
+ ResettableClass *rc = RESETTABLE_GET_CLASS(obj);
461
+ ResettableState *s = rc->get_state(obj);
462
+ const char *obj_typename = object_get_typename(obj);
463
+
464
+ assert(!s->exit_phase_in_progress);
465
+ trace_resettable_phase_exit_begin(obj, obj_typename, s->count, type);
466
+
467
+ /* exit_phase_in_progress ensures this phase is 'atomic' */
468
+ s->exit_phase_in_progress = true;
469
+ resettable_child_foreach(rc, obj, resettable_phase_exit, NULL, type);
470
+
471
+ assert(s->count > 0);
472
+ if (s->count == 1) {
473
+ trace_resettable_phase_exit_exec(obj, obj_typename, !!rc->phases.exit);
474
+ if (rc->phases.exit && !resettable_get_tr_func(rc, obj)) {
475
+ rc->phases.exit(obj);
476
+ }
477
+ s->count = 0;
478
+ }
479
+ s->exit_phase_in_progress = false;
480
+ trace_resettable_phase_exit_end(obj, obj_typename, s->count);
481
+}
482
+
483
+void resettable_class_set_parent_phases(ResettableClass *rc,
484
+ ResettableEnterPhase enter,
485
+ ResettableHoldPhase hold,
486
+ ResettableExitPhase exit,
487
+ ResettablePhases *parent_phases)
488
+{
489
+ *parent_phases = rc->phases;
490
+ if (enter) {
491
+ rc->phases.enter = enter;
492
+ }
493
+ if (hold) {
494
+ rc->phases.hold = hold;
495
+ }
496
+ if (exit) {
497
+ rc->phases.exit = exit;
498
+ }
499
+}
500
+
501
+static const TypeInfo resettable_interface_info = {
502
+ .name = TYPE_RESETTABLE_INTERFACE,
503
+ .parent = TYPE_INTERFACE,
504
+ .class_size = sizeof(ResettableClass),
505
+};
506
+
507
+static void reset_register_types(void)
508
+{
509
+ type_register_static(&resettable_interface_info);
510
+}
511
+
512
+type_init(reset_register_types)
513
diff --git a/hw/core/trace-events b/hw/core/trace-events
514
index XXXXXXX..XXXXXXX 100644
515
--- a/hw/core/trace-events
516
+++ b/hw/core/trace-events
517
@@ -XXX,XX +XXX,XX @@ qbus_reset(void *obj, const char *objtype) "obj=%p(%s)"
518
qbus_reset_all(void *obj, const char *objtype) "obj=%p(%s)"
519
qbus_reset_tree(void *obj, const char *objtype) "obj=%p(%s)"
520
qdev_update_parent_bus(void *obj, const char *objtype, void *oldp, const char *oldptype, void *newp, const char *newptype) "obj=%p(%s) old_parent=%p(%s) new_parent=%p(%s)"
521
+
522
+# resettable.c
523
+resettable_reset(void *obj, int cold) "obj=%p cold=%d"
524
+resettable_reset_assert_begin(void *obj, int cold) "obj=%p cold=%d"
525
+resettable_reset_assert_end(void *obj) "obj=%p"
526
+resettable_reset_release_begin(void *obj, int cold) "obj=%p cold=%d"
527
+resettable_reset_release_end(void *obj) "obj=%p"
528
+resettable_phase_enter_begin(void *obj, const char *objtype, unsigned count, int type) "obj=%p(%s) count=%d type=%d"
529
+resettable_phase_enter_exec(void *obj, const char *objtype, int type, int has_method) "obj=%p(%s) type=%d method=%d"
530
+resettable_phase_enter_end(void *obj, const char *objtype, unsigned count) "obj=%p(%s) count=%d"
531
+resettable_phase_hold_begin(void *obj, const char *objtype, unsigned count, int type) "obj=%p(%s) count=%d type=%d"
532
+resettable_phase_hold_exec(void *obj, const char *objtype, int has_method) "obj=%p(%s) method=%d"
533
+resettable_phase_hold_end(void *obj, const char *objtype, unsigned count) "obj=%p(%s) count=%d"
534
+resettable_phase_exit_begin(void *obj, const char *objtype, unsigned count, int type) "obj=%p(%s) count=%d type=%d"
535
+resettable_phase_exit_exec(void *obj, const char *objtype, int has_method) "obj=%p(%s) method=%d"
536
+resettable_phase_exit_end(void *obj, const char *objtype, unsigned count) "obj=%p(%s) count=%d"
537
+resettable_transitional_function(void *obj, const char *objtype) "obj=%p(%s)"
538
--
539
2.20.1
540
541
diff view generated by jsdifflib
New patch
1
1
From: Damien Hedde <damien.hedde@greensocs.com>
2
3
This commit adds support of Resettable interface to buses and devices:
4
+ ResettableState structure is added in the Bus/Device state
5
+ Resettable methods are implemented.
6
+ device/bus_is_in_reset function defined
7
8
This commit allows to transition the objects to the new
9
multi-phase interface without changing the reset behavior at all.
10
Object single reset method can be split into the 3 different phases
11
but the 3 phases are still executed in a row for a given object.
12
From the qdev/qbus reset api point of view, nothing is changed.
13
qdev_reset_all() and qbus_reset_all() are not modified as well as
14
device_legacy_reset().
15
16
Transition of an object must be done from parent class to child class.
17
Care has been taken to allow the transition of a parent class
18
without requiring the child classes to be transitioned at the same
19
time. Note that SysBus and SysBusDevice class do not need any transition
20
because they do not override the legacy reset method.
21
22
Signed-off-by: Damien Hedde <damien.hedde@greensocs.com>
23
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
24
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
25
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
26
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
27
Message-id: 20200123132823.1117486-5-damien.hedde@greensocs.com
28
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
29
---
30
tests/Makefile.include | 1 +
31
include/hw/qdev-core.h | 27 ++++++++++++
32
hw/core/bus.c | 97 ++++++++++++++++++++++++++++++++++++++++++
33
hw/core/qdev.c | 93 ++++++++++++++++++++++++++++++++++++++++
34
4 files changed, 218 insertions(+)
35
36
diff --git a/tests/Makefile.include b/tests/Makefile.include
37
index XXXXXXX..XXXXXXX 100644
38
--- a/tests/Makefile.include
39
+++ b/tests/Makefile.include
40
@@ -XXX,XX +XXX,XX @@ tests/fp/%:
41
tests/test-qdev-global-props$(EXESUF): tests/test-qdev-global-props.o \
42
    hw/core/qdev.o hw/core/qdev-properties.o hw/core/hotplug.o\
43
    hw/core/bus.o \
44
+    hw/core/resettable.o \
45
    hw/core/irq.o \
46
    hw/core/fw-path-provider.o \
47
    hw/core/reset.o \
48
diff --git a/include/hw/qdev-core.h b/include/hw/qdev-core.h
49
index XXXXXXX..XXXXXXX 100644
50
--- a/include/hw/qdev-core.h
51
+++ b/include/hw/qdev-core.h
52
@@ -XXX,XX +XXX,XX @@
53
#include "qemu/bitmap.h"
54
#include "qom/object.h"
55
#include "hw/hotplug.h"
56
+#include "hw/resettable.h"
57
58
enum {
59
DEV_NVECTORS_UNSPECIFIED = -1,
60
@@ -XXX,XX +XXX,XX @@ typedef struct DeviceClass {
61
bool hotpluggable;
62
63
/* callbacks */
64
+ /*
65
+ * Reset method here is deprecated and replaced by methods in the
66
+ * resettable class interface to implement a multi-phase reset.
67
+ * TODO: remove once every reset callback is unused
68
+ */
69
DeviceReset reset;
70
DeviceRealize realize;
71
DeviceUnrealize unrealize;
72
@@ -XXX,XX +XXX,XX @@ struct NamedGPIOList {
73
/**
74
* DeviceState:
75
* @realized: Indicates whether the device has been fully constructed.
76
+ * @reset: ResettableState for the device; handled by Resettable interface.
77
*
78
* This structure should not be accessed directly. We declare it here
79
* so that it can be embedded in individual device state structures.
80
@@ -XXX,XX +XXX,XX @@ struct DeviceState {
81
int num_child_bus;
82
int instance_id_alias;
83
int alias_required_for_version;
84
+ ResettableState reset;
85
};
86
87
struct DeviceListener {
88
@@ -XXX,XX +XXX,XX @@ typedef struct BusChild {
89
/**
90
* BusState:
91
* @hotplug_handler: link to a hotplug handler associated with bus.
92
+ * @reset: ResettableState for the bus; handled by Resettable interface.
93
*/
94
struct BusState {
95
Object obj;
96
@@ -XXX,XX +XXX,XX @@ struct BusState {
97
int num_children;
98
QTAILQ_HEAD(, BusChild) children;
99
QLIST_ENTRY(BusState) sibling;
100
+ ResettableState reset;
101
};
102
103
/**
104
@@ -XXX,XX +XXX,XX @@ void qdev_reset_all_fn(void *opaque);
105
void qbus_reset_all(BusState *bus);
106
void qbus_reset_all_fn(void *opaque);
107
108
+/**
109
+ * device_is_in_reset:
110
+ * Return true if the device @dev is currently being reset.
111
+ */
112
+bool device_is_in_reset(DeviceState *dev);
113
+
114
+/**
115
+ * bus_is_in_reset:
116
+ * Return true if the bus @bus is currently being reset.
117
+ */
118
+bool bus_is_in_reset(BusState *bus);
119
+
120
/* This should go away once we get rid of the NULL bus hack */
121
BusState *sysbus_get_default(void);
122
123
@@ -XXX,XX +XXX,XX @@ void device_legacy_reset(DeviceState *dev);
124
125
void device_class_set_props(DeviceClass *dc, Property *props);
126
127
+/**
128
+ * device_class_set_parent_reset:
129
+ * TODO: remove the function when DeviceClass's reset method
130
+ * is not used anymore.
131
+ */
132
void device_class_set_parent_reset(DeviceClass *dc,
133
DeviceReset dev_reset,
134
DeviceReset *parent_reset);
135
diff --git a/hw/core/bus.c b/hw/core/bus.c
136
index XXXXXXX..XXXXXXX 100644
137
--- a/hw/core/bus.c
138
+++ b/hw/core/bus.c
139
@@ -XXX,XX +XXX,XX @@ int qbus_walk_children(BusState *bus,
140
return 0;
141
}
142
143
+bool bus_is_in_reset(BusState *bus)
144
+{
145
+ return resettable_is_in_reset(OBJECT(bus));
146
+}
147
+
148
+static ResettableState *bus_get_reset_state(Object *obj)
149
+{
150
+ BusState *bus = BUS(obj);
151
+ return &bus->reset;
152
+}
153
+
154
+static void bus_reset_child_foreach(Object *obj, ResettableChildCallback cb,
155
+ void *opaque, ResetType type)
156
+{
157
+ BusState *bus = BUS(obj);
158
+ BusChild *kid;
159
+
160
+ QTAILQ_FOREACH(kid, &bus->children, sibling) {
161
+ cb(OBJECT(kid->child), opaque, type);
162
+ }
163
+}
164
+
165
static void qbus_realize(BusState *bus, DeviceState *parent, const char *name)
166
{
167
const char *typename = object_get_typename(OBJECT(bus));
168
@@ -XXX,XX +XXX,XX @@ static char *default_bus_get_fw_dev_path(DeviceState *dev)
169
return g_strdup(object_get_typename(OBJECT(dev)));
170
}
171
172
+/**
173
+ * bus_phases_reset:
174
+ * Transition reset method for buses to allow moving
175
+ * smoothly from legacy reset method to multi-phases
176
+ */
177
+static void bus_phases_reset(BusState *bus)
178
+{
179
+ ResettableClass *rc = RESETTABLE_GET_CLASS(bus);
180
+
181
+ if (rc->phases.enter) {
182
+ rc->phases.enter(OBJECT(bus), RESET_TYPE_COLD);
183
+ }
184
+ if (rc->phases.hold) {
185
+ rc->phases.hold(OBJECT(bus));
186
+ }
187
+ if (rc->phases.exit) {
188
+ rc->phases.exit(OBJECT(bus));
189
+ }
190
+}
191
+
192
+static void bus_transitional_reset(Object *obj)
193
+{
194
+ BusClass *bc = BUS_GET_CLASS(obj);
195
+
196
+ /*
197
+ * This will call either @bus_phases_reset (for multi-phases transitioned
198
+ * buses) or a bus's specific method for not-yet transitioned buses.
199
+ * In both case, it does not reset children.
200
+ */
201
+ if (bc->reset) {
202
+ bc->reset(BUS(obj));
203
+ }
204
+}
205
+
206
+/**
207
+ * bus_get_transitional_reset:
208
+ * check if the bus's class is ready for multi-phase
209
+ */
210
+static ResettableTrFunction bus_get_transitional_reset(Object *obj)
211
+{
212
+ BusClass *dc = BUS_GET_CLASS(obj);
213
+ if (dc->reset != bus_phases_reset) {
214
+ /*
215
+ * dc->reset has been overridden by a subclass,
216
+ * the bus is not ready for multi phase yet.
217
+ */
218
+ return bus_transitional_reset;
219
+ }
220
+ return NULL;
221
+}
222
+
223
static void bus_class_init(ObjectClass *class, void *data)
224
{
225
BusClass *bc = BUS_CLASS(class);
226
+ ResettableClass *rc = RESETTABLE_CLASS(class);
227
228
class->unparent = bus_unparent;
229
bc->get_fw_dev_path = default_bus_get_fw_dev_path;
230
+
231
+ rc->get_state = bus_get_reset_state;
232
+ rc->child_foreach = bus_reset_child_foreach;
233
+
234
+ /*
235
+ * @bus_phases_reset is put as the default reset method below, allowing
236
+ * to do the multi-phase transition from base classes to leaf classes. It
237
+ * allows a legacy-reset Bus class to extend a multi-phases-reset
238
+ * Bus class for the following reason:
239
+ * + If a base class B has been moved to multi-phase, then it does not
240
+ * override this default reset method and may have defined phase methods.
241
+ * + A child class C (extending class B) which uses
242
+ * bus_class_set_parent_reset() (or similar means) to override the
243
+ * reset method will still work as expected. @bus_phases_reset function
244
+ * will be registered as the parent reset method and effectively call
245
+ * parent reset phases.
246
+ */
247
+ bc->reset = bus_phases_reset;
248
+ rc->get_transitional_function = bus_get_transitional_reset;
249
}
250
251
static void qbus_finalize(Object *obj)
252
@@ -XXX,XX +XXX,XX @@ static const TypeInfo bus_info = {
253
.instance_init = qbus_initfn,
254
.instance_finalize = qbus_finalize,
255
.class_init = bus_class_init,
256
+ .interfaces = (InterfaceInfo[]) {
257
+ { TYPE_RESETTABLE_INTERFACE },
258
+ { }
259
+ },
260
};
261
262
static void bus_register_types(void)
263
diff --git a/hw/core/qdev.c b/hw/core/qdev.c
264
index XXXXXXX..XXXXXXX 100644
265
--- a/hw/core/qdev.c
266
+++ b/hw/core/qdev.c
267
@@ -XXX,XX +XXX,XX @@ void qbus_reset_all_fn(void *opaque)
268
qbus_reset_all(bus);
269
}
270
271
+bool device_is_in_reset(DeviceState *dev)
272
+{
273
+ return resettable_is_in_reset(OBJECT(dev));
274
+}
275
+
276
+static ResettableState *device_get_reset_state(Object *obj)
277
+{
278
+ DeviceState *dev = DEVICE(obj);
279
+ return &dev->reset;
280
+}
281
+
282
+static void device_reset_child_foreach(Object *obj, ResettableChildCallback cb,
283
+ void *opaque, ResetType type)
284
+{
285
+ DeviceState *dev = DEVICE(obj);
286
+ BusState *bus;
287
+
288
+ QLIST_FOREACH(bus, &dev->child_bus, sibling) {
289
+ cb(OBJECT(bus), opaque, type);
290
+ }
291
+}
292
+
293
/* can be used as ->unplug() callback for the simple cases */
294
void qdev_simple_device_unplug_cb(HotplugHandler *hotplug_dev,
295
DeviceState *dev, Error **errp)
296
@@ -XXX,XX +XXX,XX @@ device_vmstate_if_get_id(VMStateIf *obj)
297
return qdev_get_dev_path(dev);
298
}
299
300
+/**
301
+ * device_phases_reset:
302
+ * Transition reset method for devices to allow moving
303
+ * smoothly from legacy reset method to multi-phases
304
+ */
305
+static void device_phases_reset(DeviceState *dev)
306
+{
307
+ ResettableClass *rc = RESETTABLE_GET_CLASS(dev);
308
+
309
+ if (rc->phases.enter) {
310
+ rc->phases.enter(OBJECT(dev), RESET_TYPE_COLD);
311
+ }
312
+ if (rc->phases.hold) {
313
+ rc->phases.hold(OBJECT(dev));
314
+ }
315
+ if (rc->phases.exit) {
316
+ rc->phases.exit(OBJECT(dev));
317
+ }
318
+}
319
+
320
+static void device_transitional_reset(Object *obj)
321
+{
322
+ DeviceClass *dc = DEVICE_GET_CLASS(obj);
323
+
324
+ /*
325
+ * This will call either @device_phases_reset (for multi-phases transitioned
326
+ * devices) or a device's specific method for not-yet transitioned devices.
327
+ * In both case, it does not reset children.
328
+ */
329
+ if (dc->reset) {
330
+ dc->reset(DEVICE(obj));
331
+ }
332
+}
333
+
334
+/**
335
+ * device_get_transitional_reset:
336
+ * check if the device's class is ready for multi-phase
337
+ */
338
+static ResettableTrFunction device_get_transitional_reset(Object *obj)
339
+{
340
+ DeviceClass *dc = DEVICE_GET_CLASS(obj);
341
+ if (dc->reset != device_phases_reset) {
342
+ /*
343
+ * dc->reset has been overridden by a subclass,
344
+ * the device is not ready for multi phase yet.
345
+ */
346
+ return device_transitional_reset;
347
+ }
348
+ return NULL;
349
+}
350
+
351
static void device_class_init(ObjectClass *class, void *data)
352
{
353
DeviceClass *dc = DEVICE_CLASS(class);
354
VMStateIfClass *vc = VMSTATE_IF_CLASS(class);
355
+ ResettableClass *rc = RESETTABLE_CLASS(class);
356
357
class->unparent = device_unparent;
358
359
@@ -XXX,XX +XXX,XX @@ static void device_class_init(ObjectClass *class, void *data)
360
dc->hotpluggable = true;
361
dc->user_creatable = true;
362
vc->get_id = device_vmstate_if_get_id;
363
+ rc->get_state = device_get_reset_state;
364
+ rc->child_foreach = device_reset_child_foreach;
365
+
366
+ /*
367
+ * @device_phases_reset is put as the default reset method below, allowing
368
+ * to do the multi-phase transition from base classes to leaf classes. It
369
+ * allows a legacy-reset Device class to extend a multi-phases-reset
370
+ * Device class for the following reason:
371
+ * + If a base class B has been moved to multi-phase, then it does not
372
+ * override this default reset method and may have defined phase methods.
373
+ * + A child class C (extending class B) which uses
374
+ * device_class_set_parent_reset() (or similar means) to override the
375
+ * reset method will still work as expected. @device_phases_reset function
376
+ * will be registered as the parent reset method and effectively call
377
+ * parent reset phases.
378
+ */
379
+ dc->reset = device_phases_reset;
380
+ rc->get_transitional_function = device_get_transitional_reset;
381
382
object_class_property_add_bool(class, "realized",
383
device_get_realized, device_set_realized,
384
@@ -XXX,XX +XXX,XX @@ static const TypeInfo device_type_info = {
385
.class_size = sizeof(DeviceClass),
386
.interfaces = (InterfaceInfo[]) {
387
{ TYPE_VMSTATE_IF },
388
+ { TYPE_RESETTABLE_INTERFACE },
389
{ }
390
}
391
};
392
--
393
2.20.1
394
395
diff view generated by jsdifflib
New patch
1
From: Damien Hedde <damien.hedde@greensocs.com>
1
2
3
Add a function resettable_change_parent() to do the required
4
plumbing when changing the parent a of Resettable object.
5
6
We need to make sure that the reset state of the object remains
7
coherent with the reset state of the new parent.
8
9
We make the 2 following hypothesis:
10
+ when an object is put in a parent under reset, the object goes in
11
reset.
12
+ when an object is removed from a parent under reset, the object
13
leaves reset.
14
15
The added function avoids any glitch if both old and new parent are
16
already in reset.
17
18
Signed-off-by: Damien Hedde <damien.hedde@greensocs.com>
19
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
20
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
21
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
22
Message-id: 20200123132823.1117486-6-damien.hedde@greensocs.com
23
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
24
---
25
include/hw/resettable.h | 16 +++++++++++
26
hw/core/resettable.c | 62 +++++++++++++++++++++++++++++++++++++++--
27
hw/core/trace-events | 1 +
28
3 files changed, 77 insertions(+), 2 deletions(-)
29
30
diff --git a/include/hw/resettable.h b/include/hw/resettable.h
31
index XXXXXXX..XXXXXXX 100644
32
--- a/include/hw/resettable.h
33
+++ b/include/hw/resettable.h
34
@@ -XXX,XX +XXX,XX @@ void resettable_release_reset(Object *obj, ResetType type);
35
*/
36
bool resettable_is_in_reset(Object *obj);
37
38
+/**
39
+ * resettable_change_parent:
40
+ * Indicate that the parent of Ressettable @obj is changing from @oldp to @newp.
41
+ * All 3 objects must implement resettable interface. @oldp or @newp may be
42
+ * NULL.
43
+ *
44
+ * This function will adapt the reset state of @obj so that it is coherent
45
+ * with the reset state of @newp. It may trigger @resettable_assert_reset()
46
+ * or @resettable_release_reset(). It will do such things only if the reset
47
+ * state of @newp and @oldp are different.
48
+ *
49
+ * When using this function during reset, it must only be called during
50
+ * a hold phase method. Calling this during enter or exit phase is an error.
51
+ */
52
+void resettable_change_parent(Object *obj, Object *newp, Object *oldp);
53
+
54
/**
55
* resettable_class_set_parent_phases:
56
*
57
diff --git a/hw/core/resettable.c b/hw/core/resettable.c
58
index XXXXXXX..XXXXXXX 100644
59
--- a/hw/core/resettable.c
60
+++ b/hw/core/resettable.c
61
@@ -XXX,XX +XXX,XX @@ static void resettable_phase_exit(Object *obj, void *opaque, ResetType type);
62
* enter_phase_in_progress:
63
* True if we are currently in reset enter phase.
64
*
65
- * Note: This flag is only used to guarantee (using asserts) that the reset
66
- * API is used correctly. We can use a global variable because we rely on the
67
+ * exit_phase_in_progress:
68
+ * count the number of exit phase we are in.
69
+ *
70
+ * Note: These flags are only used to guarantee (using asserts) that the reset
71
+ * API is used correctly. We can use global variables because we rely on the
72
* iothread mutex to ensure only one reset operation is in a progress at a
73
* given time.
74
*/
75
static bool enter_phase_in_progress;
76
+static unsigned exit_phase_in_progress;
77
78
void resettable_reset(Object *obj, ResetType type)
79
{
80
@@ -XXX,XX +XXX,XX @@ void resettable_release_reset(Object *obj, ResetType type)
81
trace_resettable_reset_release_begin(obj, type);
82
assert(!enter_phase_in_progress);
83
84
+ exit_phase_in_progress += 1;
85
resettable_phase_exit(obj, NULL, type);
86
+ exit_phase_in_progress -= 1;
87
88
trace_resettable_reset_release_end(obj);
89
}
90
@@ -XXX,XX +XXX,XX @@ static void resettable_phase_exit(Object *obj, void *opaque, ResetType type)
91
trace_resettable_phase_exit_end(obj, obj_typename, s->count);
92
}
93
94
+/*
95
+ * resettable_get_count:
96
+ * Get the count of the Resettable object @obj. Return 0 if @obj is NULL.
97
+ */
98
+static unsigned resettable_get_count(Object *obj)
99
+{
100
+ if (obj) {
101
+ ResettableClass *rc = RESETTABLE_GET_CLASS(obj);
102
+ return rc->get_state(obj)->count;
103
+ }
104
+ return 0;
105
+}
106
+
107
+void resettable_change_parent(Object *obj, Object *newp, Object *oldp)
108
+{
109
+ ResettableClass *rc = RESETTABLE_GET_CLASS(obj);
110
+ ResettableState *s = rc->get_state(obj);
111
+ unsigned newp_count = resettable_get_count(newp);
112
+ unsigned oldp_count = resettable_get_count(oldp);
113
+
114
+ /*
115
+ * Ensure we do not change parent when in enter or exit phase.
116
+ * During these phases, the reset subtree being updated is partly in reset
117
+ * and partly not in reset (it depends on the actual position in
118
+ * resettable_child_foreach()s). We are not able to tell in which part is a
119
+ * leaving or arriving device. Thus we cannot set the reset count of the
120
+ * moving device to the proper value.
121
+ */
122
+ assert(!enter_phase_in_progress && !exit_phase_in_progress);
123
+ trace_resettable_change_parent(obj, oldp, oldp_count, newp, newp_count);
124
+
125
+ /*
126
+ * At most one of the two 'for' loops will be executed below
127
+ * in order to cope with the difference between the two counts.
128
+ */
129
+ /* if newp is more reset than oldp */
130
+ for (unsigned i = oldp_count; i < newp_count; i++) {
131
+ resettable_assert_reset(obj, RESET_TYPE_COLD);
132
+ }
133
+ /*
134
+ * if obj is leaving a bus under reset, we need to ensure
135
+ * hold phase is not pending.
136
+ */
137
+ if (oldp_count && s->hold_phase_pending) {
138
+ resettable_phase_hold(obj, NULL, RESET_TYPE_COLD);
139
+ }
140
+ /* if oldp is more reset than newp */
141
+ for (unsigned i = newp_count; i < oldp_count; i++) {
142
+ resettable_release_reset(obj, RESET_TYPE_COLD);
143
+ }
144
+}
145
+
146
void resettable_class_set_parent_phases(ResettableClass *rc,
147
ResettableEnterPhase enter,
148
ResettableHoldPhase hold,
149
diff --git a/hw/core/trace-events b/hw/core/trace-events
150
index XXXXXXX..XXXXXXX 100644
151
--- a/hw/core/trace-events
152
+++ b/hw/core/trace-events
153
@@ -XXX,XX +XXX,XX @@ resettable_reset_assert_begin(void *obj, int cold) "obj=%p cold=%d"
154
resettable_reset_assert_end(void *obj) "obj=%p"
155
resettable_reset_release_begin(void *obj, int cold) "obj=%p cold=%d"
156
resettable_reset_release_end(void *obj) "obj=%p"
157
+resettable_change_parent(void *obj, void *o, unsigned oc, void *n, unsigned nc) "obj=%p from=%p(%d) to=%p(%d)"
158
resettable_phase_enter_begin(void *obj, const char *objtype, unsigned count, int type) "obj=%p(%s) count=%d type=%d"
159
resettable_phase_enter_exec(void *obj, const char *objtype, int type, int has_method) "obj=%p(%s) type=%d method=%d"
160
resettable_phase_enter_end(void *obj, const char *objtype, unsigned count) "obj=%p(%s) count=%d"
161
--
162
2.20.1
163
164
diff view generated by jsdifflib
New patch
1
From: Damien Hedde <damien.hedde@greensocs.com>
1
2
3
In qdev_set_parent_bus(), when changing the parent bus of a
4
realized device, if the source and destination buses are not in the
5
same reset state, some adaptations are required. This patch adds
6
needed call to resettable_change_parent() to make sure a device reset
7
state stays coherent with its parent bus.
8
9
The addition is a no-op if:
10
1. the device being parented is not realized.
11
2. the device is realized, but both buses are not under reset.
12
13
Case 2 means that as long as qdev_set_parent_bus() is called
14
during the machine realization procedure (which is before the
15
machine reset so nothing is in reset), it is a no op.
16
17
There are 52 call sites of qdev_set_parent_bus(). All but one fall
18
into the no-op case:
19
+ 29 trivial calls related to virtio (in hw/{s390x,display,virtio}/
20
{vhost,virtio}-xxx.c) to set a vdev(or vgpu) composing device
21
parent bus just before realizing the same vdev(vgpu).
22
+ hw/core/qdev.c: when creating a device in qdev_try_create()
23
+ hw/core/sysbus.c: when initializing a device in the sysbus
24
+ hw/i386/amd_iommu.c: before realizing AMDVIState/pci
25
+ hw/isa/piix4.c: before realizing PIIX4State/rtc
26
+ hw/misc/auxbus.c: when creating an AUXBus
27
+ hw/misc/auxbus.c: when creating an AUXBus child
28
+ hw/misc/macio/macio.c: when initializing a MACIOState child
29
+ hw/misc/macio/macio.c: before realizing NewWorldMacIOState/pmu
30
+ hw/misc/macio/macio.c: before realizing NewWorldMacIOState/cuda
31
+ hw/net/virtio-net.c: Used for migration when using the failover
32
mechanism to migration a vfio-pci/net. It is
33
a no-op because at this point the device is
34
already on the bus.
35
+ hw/pci-host/designware.c: before realizing DesignwarePCIEHost/root
36
+ hw/pci-host/gpex.c: before realizing GPEXHost/root
37
+ hw/pci-host/prep.c: when initialiazing PREPPCIState/pci_dev
38
+ hw/pci-host/q35.c: before realizing Q35PCIHost/mch
39
+ hw/pci-host/versatile.c: when initializing PCIVPBState/pci_dev
40
+ hw/pci-host/xilinx-pcie.c: before realizing XilinxPCIEHost/root
41
+ hw/s390x/event-facility.c: when creating SCLPEventFacility/
42
TYPE_SCLP_QUIESCE
43
+ hw/s390x/event-facility.c: ditto with SCLPEventFacility/
44
TYPE_SCLP_CPU_HOTPLUG
45
+ hw/s390x/sclp.c: Not trivial because it is called on a SLCPDevice
46
just after realizing it. Ok because at this point the destination
47
bus (sysbus) is not in reset; the realize step is before the
48
machine reset.
49
+ hw/sd/core.c: Not OK. Used in sdbus_reparent_card(). See below.
50
+ hw/ssi/ssi.c: Used to put spi slave on spi bus and connect the cs
51
line in ssi_auto_connect_slave(). Ok because this function is only
52
used in realize step in hw/ssi/aspeed_smc.ci, hw/ssi/imx_spi.c,
53
hw/ssi/mss-spi.c, hw/ssi/xilinx_spi.c and hw/ssi/xilinx_spips.c.
54
+ hw/xen/xen-legacy-backend.c: when creating a XenLegacyDevice device
55
+ qdev-monitor.c: in device hotplug creation procedure before realize
56
57
Note that this commit alone will have no effect, right now there is no
58
use of resettable API to reset anything. So a bus will never be tagged
59
as in-reset by this same API.
60
61
The one place where side-effect will occurs is in hw/sd/core.c in
62
sdbus_reparent_card(). This function is only used in the raspi machines,
63
including during the sysbus reset procedure. This case will be
64
carrefully handled when doing the multiple phase reset transition.
65
66
Signed-off-by: Damien Hedde <damien.hedde@greensocs.com>
67
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
68
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
69
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
70
Message-id: 20200123132823.1117486-7-damien.hedde@greensocs.com
71
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
72
---
73
hw/core/qdev.c | 16 +++++++++++-----
74
1 file changed, 11 insertions(+), 5 deletions(-)
75
76
diff --git a/hw/core/qdev.c b/hw/core/qdev.c
77
index XXXXXXX..XXXXXXX 100644
78
--- a/hw/core/qdev.c
79
+++ b/hw/core/qdev.c
80
@@ -XXX,XX +XXX,XX @@ static void bus_add_child(BusState *bus, DeviceState *child)
81
82
void qdev_set_parent_bus(DeviceState *dev, BusState *bus)
83
{
84
- bool replugging = dev->parent_bus != NULL;
85
+ BusState *old_parent_bus = dev->parent_bus;
86
87
- if (replugging) {
88
+ if (old_parent_bus) {
89
trace_qdev_update_parent_bus(dev, object_get_typename(OBJECT(dev)),
90
- dev->parent_bus, object_get_typename(OBJECT(dev->parent_bus)),
91
+ old_parent_bus, object_get_typename(OBJECT(old_parent_bus)),
92
OBJECT(bus), object_get_typename(OBJECT(bus)));
93
/*
94
* Keep a reference to the device while it's not plugged into
95
* any bus, to avoid it potentially evaporating when it is
96
* dereffed in bus_remove_child().
97
+ * Also keep the ref of the parent bus until the end, so that
98
+ * we can safely call resettable_change_parent() below.
99
*/
100
object_ref(OBJECT(dev));
101
bus_remove_child(dev->parent_bus, dev);
102
- object_unref(OBJECT(dev->parent_bus));
103
}
104
dev->parent_bus = bus;
105
object_ref(OBJECT(bus));
106
bus_add_child(bus, dev);
107
- if (replugging) {
108
+ if (dev->realized) {
109
+ resettable_change_parent(OBJECT(dev), OBJECT(bus),
110
+ OBJECT(old_parent_bus));
111
+ }
112
+ if (old_parent_bus) {
113
+ object_unref(OBJECT(old_parent_bus));
114
object_unref(OBJECT(dev));
115
}
116
}
117
--
118
2.20.1
119
120
diff view generated by jsdifflib
New patch
1
From: Damien Hedde <damien.hedde@greensocs.com>
1
2
3
This commit make use of the resettable API to reset the device being
4
hotplugged when it is realized. Also it ensures it is put in a reset
5
state coherent with the parent it is plugged into.
6
7
Note that there is a difference in the reset. Instead of resetting
8
only the hotplugged device, we reset also its subtree (switch to
9
resettable API). This is not expected to be a problem because
10
sub-buses are just realized too. If a hotplugged device has any
11
sub-buses it is logical to reset them too at this point.
12
13
The recently added should_be_hidden and PCI's partially_hotplugged
14
mechanisms do not interfere with realize operation:
15
+ In the should_be_hidden use case, device creation is
16
delayed.
17
+ The partially_hotplugged mechanism prevents a device to be
18
unplugged and unrealized from qdev POV and unrealized.
19
20
Signed-off-by: Damien Hedde <damien.hedde@greensocs.com>
21
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
22
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
23
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
24
Message-id: 20200123132823.1117486-8-damien.hedde@greensocs.com
25
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
26
---
27
include/hw/resettable.h | 11 +++++++++++
28
hw/core/qdev.c | 15 ++++++++++++++-
29
2 files changed, 25 insertions(+), 1 deletion(-)
30
31
diff --git a/include/hw/resettable.h b/include/hw/resettable.h
32
index XXXXXXX..XXXXXXX 100644
33
--- a/include/hw/resettable.h
34
+++ b/include/hw/resettable.h
35
@@ -XXX,XX +XXX,XX @@ struct ResettableState {
36
bool exit_phase_in_progress;
37
};
38
39
+/**
40
+ * resettable_state_clear:
41
+ * Clear the state. It puts the state to the initial (zeroed) state required
42
+ * to reuse an object. Typically used in realize step of base classes
43
+ * implementing the interface.
44
+ */
45
+static inline void resettable_state_clear(ResettableState *state)
46
+{
47
+ memset(state, 0, sizeof(ResettableState));
48
+}
49
+
50
/**
51
* resettable_reset:
52
* Trigger a reset on an object @obj of type @type. @obj must implement
53
diff --git a/hw/core/qdev.c b/hw/core/qdev.c
54
index XXXXXXX..XXXXXXX 100644
55
--- a/hw/core/qdev.c
56
+++ b/hw/core/qdev.c
57
@@ -XXX,XX +XXX,XX @@ static void device_set_realized(Object *obj, bool value, Error **errp)
58
}
59
}
60
61
+ /*
62
+ * Clear the reset state, in case the object was previously unrealized
63
+ * with a dirty state.
64
+ */
65
+ resettable_state_clear(&dev->reset);
66
+
67
QLIST_FOREACH(bus, &dev->child_bus, sibling) {
68
object_property_set_bool(OBJECT(bus), true, "realized",
69
&local_err);
70
@@ -XXX,XX +XXX,XX @@ static void device_set_realized(Object *obj, bool value, Error **errp)
71
}
72
}
73
if (dev->hotplugged) {
74
- device_legacy_reset(dev);
75
+ /*
76
+ * Reset the device, as well as its subtree which, at this point,
77
+ * should be realized too.
78
+ */
79
+ resettable_assert_reset(OBJECT(dev), RESET_TYPE_COLD);
80
+ resettable_change_parent(OBJECT(dev), OBJECT(dev->parent_bus),
81
+ NULL);
82
+ resettable_release_reset(OBJECT(dev), RESET_TYPE_COLD);
83
}
84
dev->pending_deleted_event = false;
85
86
--
87
2.20.1
88
89
diff view generated by jsdifflib
New patch
1
From: Damien Hedde <damien.hedde@greensocs.com>
1
2
3
Deprecate device_legacy_reset(), qdev_reset_all() and
4
qbus_reset_all() to be replaced by new functions
5
device_cold_reset() and bus_cold_reset() which uses resettable API.
6
7
Also introduce resettable_cold_reset_fn() which may be used as a
8
replacement for qdev_reset_all_fn and qbus_reset_all_fn().
9
10
Following patches will be needed to look at legacy reset call sites
11
and switch to resettable api. The legacy functions will be removed
12
when unused.
13
14
Signed-off-by: Damien Hedde <damien.hedde@greensocs.com>
15
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
16
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
17
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
18
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
19
Message-id: 20200123132823.1117486-9-damien.hedde@greensocs.com
20
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
21
---
22
include/hw/qdev-core.h | 27 +++++++++++++++++++++++++++
23
include/hw/resettable.h | 9 +++++++++
24
hw/core/bus.c | 5 +++++
25
hw/core/qdev.c | 5 +++++
26
hw/core/resettable.c | 5 +++++
27
5 files changed, 51 insertions(+)
28
29
diff --git a/include/hw/qdev-core.h b/include/hw/qdev-core.h
30
index XXXXXXX..XXXXXXX 100644
31
--- a/include/hw/qdev-core.h
32
+++ b/include/hw/qdev-core.h
33
@@ -XXX,XX +XXX,XX @@ int qdev_walk_children(DeviceState *dev,
34
qdev_walkerfn *post_devfn, qbus_walkerfn *post_busfn,
35
void *opaque);
36
37
+/**
38
+ * @qdev_reset_all:
39
+ * Reset @dev. See @qbus_reset_all() for more details.
40
+ *
41
+ * Note: This function is deprecated and will be removed when it becomes unused.
42
+ * Please use device_cold_reset() now.
43
+ */
44
void qdev_reset_all(DeviceState *dev);
45
void qdev_reset_all_fn(void *opaque);
46
47
@@ -XXX,XX +XXX,XX @@ void qdev_reset_all_fn(void *opaque);
48
* hard reset means that qbus_reset_all will reset all state of the device.
49
* For PCI devices, for example, this will include the base address registers
50
* or configuration space.
51
+ *
52
+ * Note: This function is deprecated and will be removed when it becomes unused.
53
+ * Please use bus_cold_reset() now.
54
*/
55
void qbus_reset_all(BusState *bus);
56
void qbus_reset_all_fn(void *opaque);
57
58
+/**
59
+ * device_cold_reset:
60
+ * Reset device @dev and perform a recursive processing using the resettable
61
+ * interface. It triggers a RESET_TYPE_COLD.
62
+ */
63
+void device_cold_reset(DeviceState *dev);
64
+
65
+/**
66
+ * bus_cold_reset:
67
+ *
68
+ * Reset bus @bus and perform a recursive processing using the resettable
69
+ * interface. It triggers a RESET_TYPE_COLD.
70
+ */
71
+void bus_cold_reset(BusState *bus);
72
+
73
/**
74
* device_is_in_reset:
75
* Return true if the device @dev is currently being reset.
76
@@ -XXX,XX +XXX,XX @@ void qdev_machine_init(void);
77
* device_legacy_reset:
78
*
79
* Reset a single device (by calling the reset method).
80
+ * Note: This function is deprecated and will be removed when it becomes unused.
81
+ * Please use device_cold_reset() now.
82
*/
83
void device_legacy_reset(DeviceState *dev);
84
85
diff --git a/include/hw/resettable.h b/include/hw/resettable.h
86
index XXXXXXX..XXXXXXX 100644
87
--- a/include/hw/resettable.h
88
+++ b/include/hw/resettable.h
89
@@ -XXX,XX +XXX,XX @@ bool resettable_is_in_reset(Object *obj);
90
*/
91
void resettable_change_parent(Object *obj, Object *newp, Object *oldp);
92
93
+/**
94
+ * resettable_cold_reset_fn:
95
+ * Helper to call resettable_reset((Object *) opaque, RESET_TYPE_COLD).
96
+ *
97
+ * This function is typically useful to register a reset handler with
98
+ * qemu_register_reset.
99
+ */
100
+void resettable_cold_reset_fn(void *opaque);
101
+
102
/**
103
* resettable_class_set_parent_phases:
104
*
105
diff --git a/hw/core/bus.c b/hw/core/bus.c
106
index XXXXXXX..XXXXXXX 100644
107
--- a/hw/core/bus.c
108
+++ b/hw/core/bus.c
109
@@ -XXX,XX +XXX,XX @@ int qbus_walk_children(BusState *bus,
110
return 0;
111
}
112
113
+void bus_cold_reset(BusState *bus)
114
+{
115
+ resettable_reset(OBJECT(bus), RESET_TYPE_COLD);
116
+}
117
+
118
bool bus_is_in_reset(BusState *bus)
119
{
120
return resettable_is_in_reset(OBJECT(bus));
121
diff --git a/hw/core/qdev.c b/hw/core/qdev.c
122
index XXXXXXX..XXXXXXX 100644
123
--- a/hw/core/qdev.c
124
+++ b/hw/core/qdev.c
125
@@ -XXX,XX +XXX,XX @@ void qbus_reset_all_fn(void *opaque)
126
qbus_reset_all(bus);
127
}
128
129
+void device_cold_reset(DeviceState *dev)
130
+{
131
+ resettable_reset(OBJECT(dev), RESET_TYPE_COLD);
132
+}
133
+
134
bool device_is_in_reset(DeviceState *dev)
135
{
136
return resettable_is_in_reset(OBJECT(dev));
137
diff --git a/hw/core/resettable.c b/hw/core/resettable.c
138
index XXXXXXX..XXXXXXX 100644
139
--- a/hw/core/resettable.c
140
+++ b/hw/core/resettable.c
141
@@ -XXX,XX +XXX,XX @@ void resettable_change_parent(Object *obj, Object *newp, Object *oldp)
142
}
143
}
144
145
+void resettable_cold_reset_fn(void *opaque)
146
+{
147
+ resettable_reset((Object *) opaque, RESET_TYPE_COLD);
148
+}
149
+
150
void resettable_class_set_parent_phases(ResettableClass *rc,
151
ResettableEnterPhase enter,
152
ResettableHoldPhase hold,
153
--
154
2.20.1
155
156
diff view generated by jsdifflib
New patch
1
From: Damien Hedde <damien.hedde@greensocs.com>
1
2
3
Signed-off-by: Damien Hedde <damien.hedde@greensocs.com>
4
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
5
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
6
Message-id: 20200123132823.1117486-10-damien.hedde@greensocs.com
7
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
8
---
9
docs/devel/index.rst | 1 +
10
docs/devel/reset.rst | 289 +++++++++++++++++++++++++++++++++++++++++++
11
2 files changed, 290 insertions(+)
12
create mode 100644 docs/devel/reset.rst
13
14
diff --git a/docs/devel/index.rst b/docs/devel/index.rst
15
index XXXXXXX..XXXXXXX 100644
16
--- a/docs/devel/index.rst
17
+++ b/docs/devel/index.rst
18
@@ -XXX,XX +XXX,XX @@ Contents:
19
tcg
20
tcg-plugins
21
bitops
22
+ reset
23
diff --git a/docs/devel/reset.rst b/docs/devel/reset.rst
24
new file mode 100644
25
index XXXXXXX..XXXXXXX
26
--- /dev/null
27
+++ b/docs/devel/reset.rst
28
@@ -XXX,XX +XXX,XX @@
29
+
30
+=======================================
31
+Reset in QEMU: the Resettable interface
32
+=======================================
33
+
34
+The reset of qemu objects is handled using the resettable interface declared
35
+in ``include/hw/resettable.h``.
36
+
37
+This interface allows objects to be grouped (on a tree basis); so that the
38
+whole group can be reset consistently. Each individual member object does not
39
+have to care about others; in particular, problems of order (which object is
40
+reset first) are addressed.
41
+
42
+As of now DeviceClass and BusClass implement this interface.
43
+
44
+
45
+Triggering reset
46
+----------------
47
+
48
+This section documents the APIs which "users" of a resettable object should use
49
+to control it. All resettable control functions must be called while holding
50
+the iothread lock.
51
+
52
+You can apply a reset to an object using ``resettable_assert_reset()``. You need
53
+to call ``resettable_release_reset()`` to release the object from reset. To
54
+instantly reset an object, without keeping it in reset state, just call
55
+``resettable_reset()``. These functions take two parameters: a pointer to the
56
+object to reset and a reset type.
57
+
58
+Several types of reset will be supported. For now only cold reset is defined;
59
+others may be added later. The Resettable interface handles reset types with an
60
+enum:
61
+
62
+``RESET_TYPE_COLD``
63
+ Cold reset is supported by every resettable object. In QEMU, it means we reset
64
+ to the initial state corresponding to the start of QEMU; this might differ
65
+ from what is a real hardware cold reset. It differs from other resets (like
66
+ warm or bus resets) which may keep certain parts untouched.
67
+
68
+Calling ``resettable_reset()`` is equivalent to calling
69
+``resettable_assert_reset()`` then ``resettable_release_reset()``. It is
70
+possible to interleave multiple calls to these three functions. There may
71
+be several reset sources/controllers of a given object. The interface handles
72
+everything and the different reset controllers do not need to know anything
73
+about each others. The object will leave reset state only when each other
74
+controllers end their reset operation. This point is handled internally by
75
+maintaining a count of in-progress resets; it is crucial to call
76
+``resettable_release_reset()`` one time and only one time per
77
+``resettable_assert_reset()`` call.
78
+
79
+For now migration of a device or bus in reset is not supported. Care must be
80
+taken not to delay ``resettable_release_reset()`` after its
81
+``resettable_assert_reset()`` counterpart.
82
+
83
+Note that, since resettable is an interface, the API takes a simple Object as
84
+parameter. Still, it is a programming error to call a resettable function on a
85
+non-resettable object and it will trigger a run time assert error. Since most
86
+calls to resettable interface are done through base class functions, such an
87
+error is not likely to happen.
88
+
89
+For Devices and Buses, the following helper functions exist:
90
+
91
+- ``device_cold_reset()``
92
+- ``bus_cold_reset()``
93
+
94
+These are simple wrappers around resettable_reset() function; they only cast the
95
+Device or Bus into an Object and pass the cold reset type. When possible
96
+prefer to use these functions instead of ``resettable_reset()``.
97
+
98
+Device and bus functions co-exist because there can be semantic differences
99
+between resetting a bus and resetting the controller bridge which owns it.
100
+For example, consider a SCSI controller. Resetting the controller puts all
101
+its registers back to what reset state was as well as reset everything on the
102
+SCSI bus, whereas resetting just the SCSI bus only resets everything that's on
103
+it but not the controller.
104
+
105
+
106
+Multi-phase mechanism
107
+---------------------
108
+
109
+This section documents the internals of the resettable interface.
110
+
111
+The resettable interface uses a multi-phase system to relieve objects and
112
+machines from reset ordering problems. To address this, the reset operation
113
+of an object is split into three well defined phases.
114
+
115
+When resetting several objects (for example the whole machine at simulation
116
+startup), all first phases of all objects are executed, then all second phases
117
+and then all third phases.
118
+
119
+The three phases are:
120
+
121
+1. The **enter** phase is executed when the object enters reset. It resets only
122
+ local state of the object; it must not do anything that has a side-effect
123
+ on other objects, such as raising or lowering a qemu_irq line or reading or
124
+ writing guest memory.
125
+
126
+2. The **hold** phase is executed for entry into reset, once every object in the
127
+ group which is being reset has had its *enter* phase executed. At this point
128
+ devices can do actions that affect other objects.
129
+
130
+3. The **exit** phase is executed when the object leaves the reset state.
131
+ Actions affecting other objects are permitted.
132
+
133
+As said in previous section, the interface maintains a count of reset. This
134
+count is used to ensure phases are executed only when required. *enter* and
135
+*hold* phases are executed only when asserting reset for the first time
136
+(if an object is already in reset state when calling
137
+``resettable_assert_reset()`` or ``resettable_reset()``, they are not
138
+executed).
139
+The *exit* phase is executed only when the last reset operation ends. Therefore
140
+the object does not need to care how many of reset controllers it has and how
141
+many of them have started a reset.
142
+
143
+
144
+Handling reset in a resettable object
145
+-------------------------------------
146
+
147
+This section documents the APIs that an implementation of a resettable object
148
+must provide and what functions it has access to. It is intended for people
149
+who want to implement or convert a class which has the resettable interface;
150
+for example when specializing an existing device or bus.
151
+
152
+Methods to implement
153
+....................
154
+
155
+Three methods should be defined or left empty. Each method corresponds to a
156
+phase of the reset; they are name ``phases.enter()``, ``phases.hold()`` and
157
+``phases.exit()``. They all take the object as parameter. The *enter* method
158
+also take the reset type as second parameter.
159
+
160
+When extending an existing class, these methods may need to be extended too.
161
+The ``resettable_class_set_parent_phases()`` class function may be used to
162
+backup parent class methods.
163
+
164
+Here follows an example to implement reset for a Device which sets an IO while
165
+in reset.
166
+
167
+::
168
+
169
+ static void mydev_reset_enter(Object *obj, ResetType type)
170
+ {
171
+ MyDevClass *myclass = MYDEV_GET_CLASS(obj);
172
+ MyDevState *mydev = MYDEV(obj);
173
+ /* call parent class enter phase */
174
+ if (myclass->parent_phases.enter) {
175
+ myclass->parent_phases.enter(obj, type);
176
+ }
177
+ /* initialize local state only */
178
+ mydev->var = 0;
179
+ }
180
+
181
+ static void mydev_reset_hold(Object *obj)
182
+ {
183
+ MyDevClass *myclass = MYDEV_GET_CLASS(obj);
184
+ MyDevState *mydev = MYDEV(obj);
185
+ /* call parent class hold phase */
186
+ if (myclass->parent_phases.hold) {
187
+ myclass->parent_phases.hold(obj);
188
+ }
189
+ /* set an IO */
190
+ qemu_set_irq(mydev->irq, 1);
191
+ }
192
+
193
+ static void mydev_reset_exit(Object *obj)
194
+ {
195
+ MyDevClass *myclass = MYDEV_GET_CLASS(obj);
196
+ MyDevState *mydev = MYDEV(obj);
197
+ /* call parent class exit phase */
198
+ if (myclass->parent_phases.exit) {
199
+ myclass->parent_phases.exit(obj);
200
+ }
201
+ /* clear an IO */
202
+ qemu_set_irq(mydev->irq, 0);
203
+ }
204
+
205
+ typedef struct MyDevClass {
206
+ MyParentClass parent_class;
207
+ /* to store eventual parent reset methods */
208
+ ResettablePhases parent_phases;
209
+ } MyDevClass;
210
+
211
+ static void mydev_class_init(ObjectClass *class, void *data)
212
+ {
213
+ MyDevClass *myclass = MYDEV_CLASS(class);
214
+ ResettableClass *rc = RESETTABLE_CLASS(class);
215
+ resettable_class_set_parent_reset_phases(rc,
216
+ mydev_reset_enter,
217
+ mydev_reset_hold,
218
+ mydev_reset_exit,
219
+ &myclass->parent_phases);
220
+ }
221
+
222
+In the above example, we override all three phases. It is possible to override
223
+only some of them by passing NULL instead of a function pointer to
224
+``resettable_class_set_parent_reset_phases()``. For example, the following will
225
+only override the *enter* phase and leave *hold* and *exit* untouched::
226
+
227
+ resettable_class_set_parent_reset_phases(rc, mydev_reset_enter,
228
+ NULL, NULL,
229
+ &myclass->parent_phases);
230
+
231
+This is equivalent to providing a trivial implementation of the hold and exit
232
+phases which does nothing but call the parent class's implementation of the
233
+phase.
234
+
235
+Polling the reset state
236
+.......................
237
+
238
+Resettable interface provides the ``resettable_is_in_reset()`` function.
239
+This function returns true if the object parameter is currently under reset.
240
+
241
+An object is under reset from the beginning of the *init* phase to the end of
242
+the *exit* phase. During all three phases, the function will return that the
243
+object is in reset.
244
+
245
+This function may be used if the object behavior has to be adapted
246
+while in reset state. For example if a device has an irq input,
247
+it will probably need to ignore it while in reset; then it can for
248
+example check the reset state at the beginning of the irq callback.
249
+
250
+Note that until migration of the reset state is supported, an object
251
+should not be left in reset. So apart from being currently executing
252
+one of the reset phases, the only cases when this function will return
253
+true is if an external interaction (like changing an io) is made during
254
+*hold* or *exit* phase of another object in the same reset group.
255
+
256
+Helpers ``device_is_in_reset()`` and ``bus_is_in_reset()`` are also provided
257
+for devices and buses and should be preferred.
258
+
259
+
260
+Base class handling of reset
261
+----------------------------
262
+
263
+This section documents parts of the reset mechanism that you only need to know
264
+about if you are extending it to work with a new base class other than
265
+DeviceClass or BusClass, or maintaining the existing code in those classes. Most
266
+people can ignore it.
267
+
268
+Methods to implement
269
+....................
270
+
271
+There are two other methods that need to exist in a class implementing the
272
+interface: ``get_state()`` and ``child_foreach()``.
273
+
274
+``get_state()`` is simple. *resettable* is an interface and, as a consequence,
275
+does not have any class state structure. But in order to factorize the code, we
276
+need one. This method must return a pointer to ``ResettableState`` structure.
277
+The structure must be allocated by the base class; preferably it should be
278
+located inside the object instance structure.
279
+
280
+``child_foreach()`` is more complex. It should execute the given callback on
281
+every reset child of the given resettable object. All children must be
282
+resettable too. Additional parameters (a reset type and an opaque pointer) must
283
+be passed to the callback too.
284
+
285
+In ``DeviceClass`` and ``BusClass`` the ``ResettableState`` is located
286
+``DeviceState`` and ``BusState`` structure. ``child_foreach()`` is implemented
287
+to follow the bus hierarchy; for a bus, it calls the function on every child
288
+device; for a device, it calls the function on every bus child. When we reset
289
+the main system bus, we reset the whole machine bus tree.
290
+
291
+Changing a resettable parent
292
+............................
293
+
294
+One thing which should be taken care of by the base class is handling reset
295
+hierarchy changes.
296
+
297
+The reset hierarchy is supposed to be static and built during machine creation.
298
+But there are actually some exceptions. To cope with this, the resettable API
299
+provides ``resettable_change_parent()``. This function allows to set, update or
300
+remove the parent of a resettable object after machine creation is done. As
301
+parameters, it takes the object being moved, the old parent if any and the new
302
+parent if any.
303
+
304
+This function can be used at any time when not in a reset operation. During
305
+a reset operation it must be used only in *hold* phase. Using it in *enter* or
306
+*exit* phase is an error.
307
+Also it should not be used during machine creation, although it is harmless to
308
+do so: the function is a no-op as long as old and new parent are NULL or not
309
+in reset.
310
+
311
+There is currently 2 cases where this function is used:
312
+
313
+1. *device hotplug*; it means a new device is introduced on a live bus.
314
+
315
+2. *hot bus change*; it means an existing live device is added, moved or
316
+ removed in the bus hierarchy. At the moment, it occurs only in the raspi
317
+ machines for changing the sdbus used by sd card.
318
--
319
2.20.1
320
321
diff view generated by jsdifflib
New patch
1
From: Damien Hedde <damien.hedde@greensocs.com>
1
2
3
Replace deprecated qbus_reset_all by resettable_cold_reset_fn for
4
the sysbus reset registration.
5
6
Apart for the raspi machines, this does not impact the behavior
7
because:
8
+ at this point resettable just calls the old reset methods of devices
9
and buses in the same order as qdev/qbus.
10
+ resettable handlers registered with qemu_register_reset are
11
serialized; there is no interleaving.
12
+ eventual explicit calls to legacy reset API (device_reset or
13
qdev/qbus_reset) inside this reset handler will not be masked out
14
by resettable mechanism; they do not go through resettable api.
15
16
For the raspi machines, during the sysbus reset the sd-card is not
17
reset twice anymore but only once. This is a consequence of switching
18
both sysbus reset and changing parent to resettable; it detects the
19
second reset is not needed. This has no impact on the state after
20
reset; the sd-card reset method only reset local state and query
21
information from the block backend.
22
23
The raspi reset change can be observed by using the following command
24
(reset will occurs, then do Ctrl-C to end qemu; no firmware is
25
given here).
26
qemu-system-aarch64 -M raspi3 \
27
-trace resettable_phase_hold_exec \
28
-trace qdev_update_parent_bus \
29
-trace resettable_change_parent \
30
-trace qdev_reset -trace qbus_reset
31
32
Before the patch, the qdev/qbus_reset traces show when reset method are
33
called. After the patch, the resettable_phase_hold_exec show when reset
34
method are called.
35
36
The traced reset order of the raspi3 is listed below. I've added empty
37
lines and the tree structure.
38
39
+->bcm2835-peripherals reset
40
|
41
| +->sd-card reset
42
| +->sd-bus reset
43
+->bcm2835_gpio reset
44
| -> dev_update_parent_bus (move the sd-card on the sdhci-bus)
45
| -> resettable_change_parent
46
|
47
+->bcm2835-dma reset
48
|
49
| +->bcm2835-sdhost-bus reset
50
+->bcm2835-sdhost reset
51
|
52
| +->sd-card (reset ONLY BEFORE BEFORE THE PATCH)
53
| +->sdhci-bus reset
54
+->generic-sdhci reset
55
|
56
+->bcm2835-rng reset
57
+->bcm2835-property reset
58
+->bcm2835-fb reset
59
+->bcm2835-mbox reset
60
+->bcm2835-aux reset
61
+->pl011 reset
62
+->bcm2835-ic reset
63
+->bcm2836-control reset
64
System reset
65
66
In both case, the sd-card is reset (being on bcm2835_gpio/sd-bus) then moved
67
to generic-sdhci/sdhci-bus by the bcm2835_gpio reset method.
68
69
Before the patch, it is then reset again being part of generic-sdhci/sdhci-bus.
70
After the patch, it considered again for reset but its reset method is not
71
called because it is already flagged as reset.
72
73
Signed-off-by: Damien Hedde <damien.hedde@greensocs.com>
74
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
75
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
76
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
77
Tested-by: Philippe Mathieu-Daudé <philmd@redhat.com>
78
Message-id: 20200123132823.1117486-11-damien.hedde@greensocs.com
79
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
80
---
81
vl.c | 10 +++++++++-
82
1 file changed, 9 insertions(+), 1 deletion(-)
83
84
diff --git a/vl.c b/vl.c
85
index XXXXXXX..XXXXXXX 100644
86
--- a/vl.c
87
+++ b/vl.c
88
@@ -XXX,XX +XXX,XX @@ int main(int argc, char **argv, char **envp)
89
90
/* TODO: once all bus devices are qdevified, this should be done
91
* when bus is created by qdev.c */
92
- qemu_register_reset(qbus_reset_all_fn, sysbus_get_default());
93
+ /*
94
+ * TODO: If we had a main 'reset container' that the whole system
95
+ * lived in, we could reset that using the multi-phase reset
96
+ * APIs. For the moment, we just reset the sysbus, which will cause
97
+ * all devices hanging off it (and all their child buses, recursively)
98
+ * to be reset. Note that this will *not* reset any Device objects
99
+ * which are not attached to some part of the qbus tree!
100
+ */
101
+ qemu_register_reset(resettable_cold_reset_fn, sysbus_get_default());
102
qemu_run_machine_init_done_notifiers();
103
104
if (rom_check_and_register_reset() != 0) {
105
--
106
2.20.1
107
108
diff view generated by jsdifflib
1
If we're directly booting a Linux kernel and the CPU supports both
1
From: Damien Hedde <damien.hedde@greensocs.com>
2
EL3 and EL2, we start the kernel in EL2, as it expects. We must also
3
set the SCR_EL3.HCE bit in this situation, so that the HVC
4
instruction is enabled rather than UNDEFing. Otherwise at least some
5
kernels will panic when trying to initialize KVM in the guest.
6
2
3
Replace deprecated qdev_reset_all by resettable_cold_reset_fn for
4
the ipl registration in the main reset handlers.
5
6
This does not impact the behavior for the following reasons:
7
+ at this point resettable just call the old reset methods of devices
8
and buses in the same order than qdev/qbus.
9
+ resettable handlers registered with qemu_register_reset are
10
serialized; there is no interleaving.
11
+ eventual explicit calls to legacy reset API (device_reset or
12
qdev/qbus_reset) inside this reset handler will not be masked out
13
by resettable mechanism; they do not go through resettable api.
14
15
Signed-off-by: Damien Hedde <damien.hedde@greensocs.com>
16
Reviewed-by: Philippe Mathieu-Daudé <philmd@redhat.com>
17
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
18
Reviewed-by: Cornelia Huck <cohuck@redhat.com>
19
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
20
Message-id: 20200123132823.1117486-12-damien.hedde@greensocs.com
7
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
21
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
8
Message-id: 20180313153458.26822-4-peter.maydell@linaro.org
9
---
22
---
10
hw/arm/boot.c | 5 +++++
23
hw/s390x/ipl.c | 10 +++++++++-
11
1 file changed, 5 insertions(+)
24
1 file changed, 9 insertions(+), 1 deletion(-)
12
25
13
diff --git a/hw/arm/boot.c b/hw/arm/boot.c
26
diff --git a/hw/s390x/ipl.c b/hw/s390x/ipl.c
14
index XXXXXXX..XXXXXXX 100644
27
index XXXXXXX..XXXXXXX 100644
15
--- a/hw/arm/boot.c
28
--- a/hw/s390x/ipl.c
16
+++ b/hw/arm/boot.c
29
+++ b/hw/s390x/ipl.c
17
@@ -XXX,XX +XXX,XX @@ static void do_cpu_reset(void *opaque)
30
@@ -XXX,XX +XXX,XX @@ static void s390_ipl_realize(DeviceState *dev, Error **errp)
18
assert(!info->secure_board_setup);
31
*/
19
}
32
ipl->compat_start_addr = ipl->start_addr;
20
33
ipl->compat_bios_start_addr = ipl->bios_start_addr;
21
+ if (arm_feature(env, ARM_FEATURE_EL2)) {
34
- qemu_register_reset(qdev_reset_all_fn, dev);
22
+ /* If we have EL2 then Linux expects the HVC insn to work */
35
+ /*
23
+ env->cp15.scr_el3 |= SCR_HCE;
36
+ * Because this Device is not on any bus in the qbus tree (it is
24
+ }
37
+ * not a sysbus device and it's not on some other bus like a PCI
25
+
38
+ * bus) it will not be automatically reset by the 'reset the
26
/* Set to non-secure if not a secure boot */
39
+ * sysbus' hook registered by vl.c like most devices. So we must
27
if (!info->secure_boot &&
40
+ * manually register a reset hook for it.
28
(cs != first_cpu || !info->secure_board_setup)) {
41
+ * TODO: there should be a better way to do this.
42
+ */
43
+ qemu_register_reset(resettable_cold_reset_fn, dev);
44
error:
45
error_propagate(errp, err);
46
}
29
--
47
--
30
2.16.2
48
2.20.1
31
49
32
50
diff view generated by jsdifflib
1
Add some assertions that if we're about to boot an AArch64 kernel,
1
From: Zenghui Yu <yuzenghui@huawei.com>
2
the board code has not mistakenly set either secure_boot or
3
secure_board_setup. It doesn't make sense to set secure_boot,
4
because all AArch64 kernels must be booted in non-secure mode.
5
2
6
It might in theory make sense to set secure_board_setup, but
3
If LPIs are disabled, KVM will just ignore the GICR_PENDBASER.PTZ bit when
7
we don't currently support that, because only the AArch32
4
restoring GICR_CTLR. Setting PTZ here makes littlt sense in "reduce GIC
8
bootloader[] code calls this hook; bootloader_aarch64[] does not.
5
initialization time".
9
Since we don't have a current need for this functionality, just
10
assert that we don't try to use it. If it's needed we'll add
11
it later.
12
6
7
And what's worse, PTZ is generally programmed by guest to indicate to the
8
Redistributor whether the LPI Pending table is zero when enabling LPIs.
9
If migration is triggered when the PTZ has just been cleared by guest (and
10
before enabling LPIs), we will see PTZ==1 on the destination side, which
11
is not as expected. Let's just drop this hackish userspace behavior.
12
13
Also take this chance to refine the comment a bit.
14
15
Fixes: 367b9f527bec ("hw/intc/arm_gicv3_kvm: Implement get/put functions")
16
Signed-off-by: Zenghui Yu <yuzenghui@huawei.com>
17
Message-id: 20200119133051.642-1-yuzenghui@huawei.com
18
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
13
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
19
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
14
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
15
Message-id: 20180313153458.26822-3-peter.maydell@linaro.org
16
---
20
---
17
hw/arm/boot.c | 7 +++++++
21
hw/intc/arm_gicv3_kvm.c | 11 ++++-------
18
1 file changed, 7 insertions(+)
22
1 file changed, 4 insertions(+), 7 deletions(-)
19
23
20
diff --git a/hw/arm/boot.c b/hw/arm/boot.c
24
diff --git a/hw/intc/arm_gicv3_kvm.c b/hw/intc/arm_gicv3_kvm.c
21
index XXXXXXX..XXXXXXX 100644
25
index XXXXXXX..XXXXXXX 100644
22
--- a/hw/arm/boot.c
26
--- a/hw/intc/arm_gicv3_kvm.c
23
+++ b/hw/arm/boot.c
27
+++ b/hw/intc/arm_gicv3_kvm.c
24
@@ -XXX,XX +XXX,XX @@ static void do_cpu_reset(void *opaque)
28
@@ -XXX,XX +XXX,XX @@ static void kvm_arm_gicv3_put(GICv3State *s)
25
} else {
29
kvm_gicd_access(s, GICD_CTLR, &reg, true);
26
env->pstate = PSTATE_MODE_EL1h;
30
27
}
31
if (redist_typer & GICR_TYPER_PLPIS) {
28
+ /* AArch64 kernels never boot in secure mode */
32
- /* Set base addresses before LPIs are enabled by GICR_CTLR write */
29
+ assert(!info->secure_boot);
33
+ /*
30
+ /* This hook is only supported for AArch32 currently:
34
+ * Restore base addresses before LPIs are potentially enabled by
31
+ * bootloader_aarch64[] will not call the hook, and
35
+ * GICR_CTLR write
32
+ * the code above has already dropped us into EL2 or EL1.
36
+ */
33
+ */
37
for (ncpu = 0; ncpu < s->num_cpu; ncpu++) {
34
+ assert(!info->secure_board_setup);
38
GICv3CPUState *c = &s->cpu[ncpu];
35
}
39
36
40
@@ -XXX,XX +XXX,XX @@ static void kvm_arm_gicv3_put(GICv3State *s)
37
/* Set to non-secure if not a secure boot */
41
kvm_gicr_access(s, GICR_PROPBASER + 4, ncpu, &regh, true);
42
43
reg64 = c->gicr_pendbaser;
44
- if (!(c->gicr_ctlr & GICR_CTLR_ENABLE_LPIS)) {
45
- /* Setting PTZ is advised if LPIs are disabled, to reduce
46
- * GIC initialization time.
47
- */
48
- reg64 |= GICR_PENDBASER_PTZ;
49
- }
50
regl = (uint32_t)reg64;
51
kvm_gicr_access(s, GICR_PENDBASER, ncpu, &regl, true);
52
regh = (uint32_t)(reg64 >> 32);
38
--
53
--
39
2.16.2
54
2.20.1
40
55
41
56
diff view generated by jsdifflib
1
For the rpi1 and 2 we want to boot the Linux kernel via some
1
From: Andrew Jones <drjones@redhat.com>
2
custom setup code that makes sure that the SMC instruction
3
acts as a no-op, because it's used for cache maintenance.
4
The rpi3 boots AArch64 kernels, which don't need SMC for
5
cache maintenance and always expect to be booted non-secure.
6
Don't fill in the aarch32-specific parts of the binfo struct.
7
2
3
Signed-off-by: Andrew Jones <drjones@redhat.com>
4
Message-id: 20200120101023.16030-2-drjones@redhat.com
5
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
8
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9
Reviewed-by: Andrew Baumann <Andrew.Baumann@microsoft.com>
10
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
11
Message-id: 20180313153458.26822-2-peter.maydell@linaro.org
12
---
7
---
13
hw/arm/raspi.c | 17 +++++++++++++----
8
target/arm/kvm_arm.h | 46 ++++++++++++++++++++++++++------------------
14
1 file changed, 13 insertions(+), 4 deletions(-)
9
1 file changed, 27 insertions(+), 19 deletions(-)
15
10
16
diff --git a/hw/arm/raspi.c b/hw/arm/raspi.c
11
diff --git a/target/arm/kvm_arm.h b/target/arm/kvm_arm.h
17
index XXXXXXX..XXXXXXX 100644
12
index XXXXXXX..XXXXXXX 100644
18
--- a/hw/arm/raspi.c
13
--- a/target/arm/kvm_arm.h
19
+++ b/hw/arm/raspi.c
14
+++ b/target/arm/kvm_arm.h
20
@@ -XXX,XX +XXX,XX @@ static void setup_boot(MachineState *machine, int version, size_t ram_size)
15
@@ -XXX,XX +XXX,XX @@
21
binfo.board_id = raspi_boardid[version];
16
int kvm_arm_vcpu_init(CPUState *cs);
22
binfo.ram_size = ram_size;
17
23
binfo.nb_cpus = smp_cpus;
18
/**
24
- binfo.board_setup_addr = BOARDSETUP_ADDR;
19
- * kvm_arm_vcpu_finalize
25
- binfo.write_board_setup = write_board_setup;
20
+ * kvm_arm_vcpu_finalize:
26
- binfo.secure_board_setup = true;
21
* @cs: CPUState
27
- binfo.secure_boot = true;
22
- * @feature: int
28
+
23
+ * @feature: feature to finalize
29
+ if (version <= 2) {
24
*
30
+ /* The rpi1 and 2 require some custom setup code to run in Secure
25
* Finalizes the configuration of the specified VCPU feature by
31
+ * mode before booting a kernel (to set up the SMC vectors so
26
* invoking the KVM_ARM_VCPU_FINALIZE ioctl. Features requiring
32
+ * that we get a no-op SMC; this is used by Linux to call the
27
@@ -XXX,XX +XXX,XX @@ void kvm_arm_register_device(MemoryRegion *mr, uint64_t devid, uint64_t group,
33
+ * firmware for some cache maintenance operations.
28
int kvm_arm_init_cpreg_list(ARMCPU *cpu);
34
+ * The rpi3 doesn't need this.
29
35
+ */
30
/**
36
+ binfo.board_setup_addr = BOARDSETUP_ADDR;
31
- * kvm_arm_reg_syncs_via_cpreg_list
37
+ binfo.write_board_setup = write_board_setup;
32
- * regidx: KVM register index
38
+ binfo.secure_board_setup = true;
33
+ * kvm_arm_reg_syncs_via_cpreg_list:
39
+ binfo.secure_boot = true;
34
+ * @regidx: KVM register index
40
+ }
35
*
41
36
* Return true if this KVM register should be synchronized via the
42
/* Pi2 and Pi3 requires SMP setup */
37
* cpreg list of arbitrary system registers, false if it is synchronized
43
if (version >= 2) {
38
@@ -XXX,XX +XXX,XX @@ int kvm_arm_init_cpreg_list(ARMCPU *cpu);
39
bool kvm_arm_reg_syncs_via_cpreg_list(uint64_t regidx);
40
41
/**
42
- * kvm_arm_cpreg_level
43
- * regidx: KVM register index
44
+ * kvm_arm_cpreg_level:
45
+ * @regidx: KVM register index
46
*
47
* Return the level of this coprocessor/system register. Return value is
48
* either KVM_PUT_RUNTIME_STATE, KVM_PUT_RESET_STATE, or KVM_PUT_FULL_STATE.
49
@@ -XXX,XX +XXX,XX @@ void kvm_arm_init_serror_injection(CPUState *cs);
50
* @cpu: ARMCPU
51
*
52
* Get VCPU related state from kvm.
53
+ *
54
+ * Returns: 0 if success else < 0 error code
55
*/
56
int kvm_get_vcpu_events(ARMCPU *cpu);
57
58
@@ -XXX,XX +XXX,XX @@ int kvm_get_vcpu_events(ARMCPU *cpu);
59
* @cpu: ARMCPU
60
*
61
* Put VCPU related state to kvm.
62
+ *
63
+ * Returns: 0 if success else < 0 error code
64
*/
65
int kvm_put_vcpu_events(ARMCPU *cpu);
66
67
@@ -XXX,XX +XXX,XX @@ typedef struct ARMHostCPUFeatures {
68
69
/**
70
* kvm_arm_get_host_cpu_features:
71
- * @ahcc: ARMHostCPUClass to fill in
72
+ * @ahcf: ARMHostCPUClass to fill in
73
*
74
* Probe the capabilities of the host kernel's preferred CPU and fill
75
* in the ARMHostCPUClass struct accordingly.
76
+ *
77
+ * Returns true on success and false otherwise.
78
*/
79
bool kvm_arm_get_host_cpu_features(ARMHostCPUFeatures *ahcf);
80
81
@@ -XXX,XX +XXX,XX @@ void kvm_arm_set_cpu_features_from_host(ARMCPU *cpu);
82
bool kvm_arm_aarch32_supported(CPUState *cs);
83
84
/**
85
- * bool kvm_arm_pmu_supported:
86
+ * kvm_arm_pmu_supported:
87
* @cs: CPUState
88
*
89
* Returns: true if the KVM VCPU can enable its PMU
90
@@ -XXX,XX +XXX,XX @@ bool kvm_arm_aarch32_supported(CPUState *cs);
91
bool kvm_arm_pmu_supported(CPUState *cs);
92
93
/**
94
- * bool kvm_arm_sve_supported:
95
+ * kvm_arm_sve_supported:
96
* @cs: CPUState
97
*
98
* Returns true if the KVM VCPU can enable SVE and false otherwise.
99
@@ -XXX,XX +XXX,XX @@ bool kvm_arm_pmu_supported(CPUState *cs);
100
bool kvm_arm_sve_supported(CPUState *cs);
101
102
/**
103
- * kvm_arm_get_max_vm_ipa_size - Returns the number of bits in the
104
- * IPA address space supported by KVM
105
- *
106
+ * kvm_arm_get_max_vm_ipa_size:
107
* @ms: Machine state handle
108
+ *
109
+ * Returns the number of bits in the IPA address space supported by KVM
110
*/
111
int kvm_arm_get_max_vm_ipa_size(MachineState *ms);
112
113
/**
114
- * kvm_arm_sync_mpstate_to_kvm
115
+ * kvm_arm_sync_mpstate_to_kvm:
116
* @cpu: ARMCPU
117
*
118
* If supported set the KVM MP_STATE based on QEMU's model.
119
+ *
120
+ * Returns 0 on success and -1 on failure.
121
*/
122
int kvm_arm_sync_mpstate_to_kvm(ARMCPU *cpu);
123
124
/**
125
- * kvm_arm_sync_mpstate_to_qemu
126
+ * kvm_arm_sync_mpstate_to_qemu:
127
* @cpu: ARMCPU
128
*
129
* If supported get the MP_STATE from KVM and store in QEMU's model.
130
+ *
131
+ * Returns 0 on success and aborts on failure.
132
*/
133
int kvm_arm_sync_mpstate_to_qemu(ARMCPU *cpu);
134
135
@@ -XXX,XX +XXX,XX @@ int kvm_arm_set_irq(int cpu, int irqtype, int irq, int level);
136
137
static inline void kvm_arm_set_cpu_features_from_host(ARMCPU *cpu)
138
{
139
- /* This should never actually be called in the "not KVM" case,
140
+ /*
141
+ * This should never actually be called in the "not KVM" case,
142
* but set up the fields to indicate an error anyway.
143
*/
144
cpu->kvm_target = QEMU_KVM_ARM_TARGET_NONE;
145
@@ -XXX,XX +XXX,XX @@ bool kvm_arm_handle_debug(CPUState *cs, struct kvm_debug_exit_arch *debug_exit);
146
*
147
* Return: TRUE if any hardware breakpoints in use.
148
*/
149
-
150
bool kvm_arm_hw_debug_active(CPUState *cs);
151
152
/**
153
* kvm_arm_copy_hw_debug_data:
154
- *
155
* @ptr: kvm_guest_debug_arch structure
156
*
157
* Copy the architecture specific debug registers into the
158
* kvm_guest_debug ioctl structure.
159
*/
160
struct kvm_guest_debug_arch;
161
-
162
void kvm_arm_copy_hw_debug_data(struct kvm_guest_debug_arch *ptr);
163
164
/**
165
- * its_class_name
166
+ * its_class_name:
167
*
168
* Return the ITS class name to use depending on whether KVM acceleration
169
* and KVM CAP_SIGNAL_MSI are supported
44
--
170
--
45
2.16.2
171
2.20.1
46
172
47
173
diff view generated by jsdifflib
1
From: Guenter Roeck <linux@roeck-us.net>
1
From: Andrew Jones <drjones@redhat.com>
2
2
3
The sabrelite machine model used by qemu-system-arm is based on the
3
Signed-off-by: Andrew Jones <drjones@redhat.com>
4
Freescale/NXP i.MX6Q processor. This SoC has an on-board ethernet
4
Message-id: 20200120101023.16030-3-drjones@redhat.com
5
controller which is supported in QEMU using the imx_fec.c module
6
(actually called imx.enet for this model.)
7
8
The include/hw/arm/fsm-imx6.h file defines the interrupt vectors for the
9
imx.enet device like this:
10
11
#define FSL_IMX6_ENET_MAC_1588_IRQ 118
12
#define FSL_IMX6_ENET_MAC_IRQ 119
13
14
According to https://www.nxp.com/docs/en/reference-manual/IMX6DQRM.pdf,
15
page 225, in Table 3-1. ARM Cortex A9 domain interrupt summary,
16
interrupts are as follows.
17
18
150 ENET MAC 0 IRQ
19
151 ENET MAC 0 1588 Timer interrupt
20
21
where
22
23
150 - 32 == 118
24
151 - 32 == 119
25
26
In other words, the vector definitions in the fsl-imx6.h file are reversed.
27
28
Fixing the interrupts alone causes problems with older Linux kernels:
29
The Ethernet interface will fail to probe with Linux v4.9 and earlier.
30
Linux v4.1 and earlier will crash due to a bug in Ethernet driver probe
31
error handling. This is a Linux kernel problem, not a qemu problem:
32
the Linux kernel only worked by accident since it requested both interrupts.
33
34
For backward compatibility, generate the Ethernet interrupt on both interrupt
35
lines. This was shown to work from all Linux kernel releases starting with
36
v3.16.
37
38
Link: https://bugs.launchpad.net/qemu/+bug/1753309
39
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
40
Message-id: 1520723090-22130-1-git-send-email-linux@roeck-us.net
41
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
5
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
42
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
6
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
43
---
7
---
44
include/hw/arm/fsl-imx6.h | 4 ++--
8
hw/arm/virt.c | 1 +
45
hw/net/imx_fec.c | 28 +++++++++++++++++++++++++++-
9
1 file changed, 1 insertion(+)
46
2 files changed, 29 insertions(+), 3 deletions(-)
47
10
48
diff --git a/include/hw/arm/fsl-imx6.h b/include/hw/arm/fsl-imx6.h
11
diff --git a/hw/arm/virt.c b/hw/arm/virt.c
49
index XXXXXXX..XXXXXXX 100644
12
index XXXXXXX..XXXXXXX 100644
50
--- a/include/hw/arm/fsl-imx6.h
13
--- a/hw/arm/virt.c
51
+++ b/include/hw/arm/fsl-imx6.h
14
+++ b/hw/arm/virt.c
52
@@ -XXX,XX +XXX,XX @@ typedef struct FslIMX6State {
15
@@ -XXX,XX +XXX,XX @@ DEFINE_VIRT_MACHINE_AS_LATEST(5, 0)
53
#define FSL_IMX6_HDMI_MASTER_IRQ 115
16
54
#define FSL_IMX6_HDMI_CEC_IRQ 116
17
static void virt_machine_4_2_options(MachineClass *mc)
55
#define FSL_IMX6_MLB150_LOW_IRQ 117
56
-#define FSL_IMX6_ENET_MAC_1588_IRQ 118
57
-#define FSL_IMX6_ENET_MAC_IRQ 119
58
+#define FSL_IMX6_ENET_MAC_IRQ 118
59
+#define FSL_IMX6_ENET_MAC_1588_IRQ 119
60
#define FSL_IMX6_PCIE1_IRQ 120
61
#define FSL_IMX6_PCIE2_IRQ 121
62
#define FSL_IMX6_PCIE3_IRQ 122
63
diff --git a/hw/net/imx_fec.c b/hw/net/imx_fec.c
64
index XXXXXXX..XXXXXXX 100644
65
--- a/hw/net/imx_fec.c
66
+++ b/hw/net/imx_fec.c
67
@@ -XXX,XX +XXX,XX @@ static void imx_enet_write_bd(IMXENETBufDesc *bd, dma_addr_t addr)
68
69
static void imx_eth_update(IMXFECState *s)
70
{
18
{
71
- if (s->regs[ENET_EIR] & s->regs[ENET_EIMR] & ENET_INT_TS_TIMER) {
19
+ virt_machine_5_0_options(mc);
72
+ /*
20
compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
73
+ * Previous versions of qemu had the ENET_INT_MAC and ENET_INT_TS_TIMER
21
}
74
+ * interrupts swapped. This worked with older versions of Linux (4.14
22
DEFINE_VIRT_MACHINE(4, 2)
75
+ * and older) since Linux associated both interrupt lines with Ethernet
76
+ * MAC interrupts. Specifically,
77
+ * - Linux 4.15 and later have separate interrupt handlers for the MAC and
78
+ * timer interrupts. Those versions of Linux fail with versions of QEMU
79
+ * with swapped interrupt assignments.
80
+ * - In linux 4.14, both interrupt lines were registered with the Ethernet
81
+ * MAC interrupt handler. As a result, all versions of qemu happen to
82
+ * work, though that is accidental.
83
+ * - In Linux 4.9 and older, the timer interrupt was registered directly
84
+ * with the Ethernet MAC interrupt handler. The MAC interrupt was
85
+ * redirected to a GPIO interrupt to work around erratum ERR006687.
86
+ * This was implemented using the SOC's IOMUX block. In qemu, this GPIO
87
+ * interrupt never fired since IOMUX is currently not supported in qemu.
88
+ * Linux instead received MAC interrupts on the timer interrupt.
89
+ * As a result, qemu versions with the swapped interrupt assignment work,
90
+ * albeit accidentally, but qemu versions with the correct interrupt
91
+ * assignment fail.
92
+ *
93
+ * To ensure that all versions of Linux work, generate ENET_INT_MAC
94
+ * interrrupts on both interrupt lines. This should be changed if and when
95
+ * qemu supports IOMUX.
96
+ */
97
+ if (s->regs[ENET_EIR] & s->regs[ENET_EIMR] &
98
+ (ENET_INT_MAC | ENET_INT_TS_TIMER)) {
99
qemu_set_irq(s->irq[1], 1);
100
} else {
101
qemu_set_irq(s->irq[1], 0);
102
--
23
--
103
2.16.2
24
2.20.1
104
25
105
26
diff view generated by jsdifflib
1
The bcm2837 is pretty similar to the bcm2836, but it does have
1
From: Andrew Jones <drjones@redhat.com>
2
some differences. Notably, the MPIDR affinity aff1 values it
3
sets for the CPUs are 0x0, rather than the 0xf that the bcm2836
4
uses, and if this is wrong Linux will not boot.
5
2
6
Rather than trying to have one device with properties that
3
Add the missing GENERIC_TIMER feature to kvm64 cpus.
7
configure it differently for the two cases, create two
8
separate QOM devices for the two SoCs. We use the same approach
9
as hw/arm/aspeed_soc.c and share code and have a data table
10
that might differ per-SoC. For the moment the two types don't
11
actually have different behaviour.
12
4
5
We don't currently use these registers when KVM is enabled, but it's
6
probably best we add the feature flag for consistency and potential
7
future use. There's also precedent, as we add the PMU feature flag to
8
KVM enabled guests, even though we don't use those registers either.
9
10
This change was originally posted as a hunk of a different, never
11
merged patch from Bijan Mottahedeh.
12
13
Signed-off-by: Andrew Jones <drjones@redhat.com>
14
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
15
Message-id: 20200120101023.16030-4-drjones@redhat.com
13
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
16
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
14
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
15
Message-id: 20180313153458.26822-7-peter.maydell@linaro.org
16
---
17
---
17
include/hw/arm/bcm2836.h | 19 +++++++++++++++++++
18
target/arm/kvm64.c | 1 +
18
hw/arm/bcm2836.c | 37 ++++++++++++++++++++++++++++++++-----
19
1 file changed, 1 insertion(+)
19
hw/arm/raspi.c | 3 ++-
20
3 files changed, 53 insertions(+), 6 deletions(-)
21
20
22
diff --git a/include/hw/arm/bcm2836.h b/include/hw/arm/bcm2836.h
21
diff --git a/target/arm/kvm64.c b/target/arm/kvm64.c
23
index XXXXXXX..XXXXXXX 100644
22
index XXXXXXX..XXXXXXX 100644
24
--- a/include/hw/arm/bcm2836.h
23
--- a/target/arm/kvm64.c
25
+++ b/include/hw/arm/bcm2836.h
24
+++ b/target/arm/kvm64.c
26
@@ -XXX,XX +XXX,XX @@
25
@@ -XXX,XX +XXX,XX @@ bool kvm_arm_get_host_cpu_features(ARMHostCPUFeatures *ahcf)
27
26
set_feature(&features, ARM_FEATURE_NEON);
28
#define BCM283X_NCPUS 4
27
set_feature(&features, ARM_FEATURE_AARCH64);
29
28
set_feature(&features, ARM_FEATURE_PMU);
30
+/* These type names are for specific SoCs; other than instantiating
29
+ set_feature(&features, ARM_FEATURE_GENERIC_TIMER);
31
+ * them, code using these devices should always handle them via the
30
32
+ * BCM283x base class, so they have no BCM2836(obj) etc macros.
31
ahcf->features = features;
33
+ */
34
+#define TYPE_BCM2836 "bcm2836"
35
+#define TYPE_BCM2837 "bcm2837"
36
+
37
typedef struct BCM283XState {
38
/*< private >*/
39
DeviceState parent_obj;
40
@@ -XXX,XX +XXX,XX @@ typedef struct BCM283XState {
41
BCM2835PeripheralState peripherals;
42
} BCM283XState;
43
44
+typedef struct BCM283XInfo BCM283XInfo;
45
+
46
+typedef struct BCM283XClass {
47
+ DeviceClass parent_class;
48
+ const BCM283XInfo *info;
49
+} BCM283XClass;
50
+
51
+#define BCM283X_CLASS(klass) \
52
+ OBJECT_CLASS_CHECK(BCM283XClass, (klass), TYPE_BCM283X)
53
+#define BCM283X_GET_CLASS(obj) \
54
+ OBJECT_GET_CLASS(BCM283XClass, (obj), TYPE_BCM283X)
55
+
56
#endif /* BCM2836_H */
57
diff --git a/hw/arm/bcm2836.c b/hw/arm/bcm2836.c
58
index XXXXXXX..XXXXXXX 100644
59
--- a/hw/arm/bcm2836.c
60
+++ b/hw/arm/bcm2836.c
61
@@ -XXX,XX +XXX,XX @@
62
/* "QA7" (Pi2) interrupt controller and mailboxes etc. */
63
#define BCM2836_CONTROL_BASE 0x40000000
64
65
+struct BCM283XInfo {
66
+ const char *name;
67
+};
68
+
69
+static const BCM283XInfo bcm283x_socs[] = {
70
+ {
71
+ .name = TYPE_BCM2836,
72
+ },
73
+ {
74
+ .name = TYPE_BCM2837,
75
+ },
76
+};
77
+
78
static void bcm2836_init(Object *obj)
79
{
80
BCM283XState *s = BCM283X(obj);
81
@@ -XXX,XX +XXX,XX @@ static Property bcm2836_props[] = {
82
DEFINE_PROP_END_OF_LIST()
83
};
84
85
-static void bcm2836_class_init(ObjectClass *oc, void *data)
86
+static void bcm283x_class_init(ObjectClass *oc, void *data)
87
{
88
DeviceClass *dc = DEVICE_CLASS(oc);
89
+ BCM283XClass *bc = BCM283X_CLASS(oc);
90
91
- dc->props = bcm2836_props;
92
+ bc->info = data;
93
dc->realize = bcm2836_realize;
94
+ dc->props = bcm2836_props;
95
}
96
97
-static const TypeInfo bcm2836_type_info = {
98
+static const TypeInfo bcm283x_type_info = {
99
.name = TYPE_BCM283X,
100
.parent = TYPE_DEVICE,
101
.instance_size = sizeof(BCM283XState),
102
.instance_init = bcm2836_init,
103
- .class_init = bcm2836_class_init,
104
+ .class_size = sizeof(BCM283XClass),
105
+ .abstract = true,
106
};
107
108
static void bcm2836_register_types(void)
109
{
110
- type_register_static(&bcm2836_type_info);
111
+ int i;
112
+
113
+ type_register_static(&bcm283x_type_info);
114
+ for (i = 0; i < ARRAY_SIZE(bcm283x_socs); i++) {
115
+ TypeInfo ti = {
116
+ .name = bcm283x_socs[i].name,
117
+ .parent = TYPE_BCM283X,
118
+ .class_init = bcm283x_class_init,
119
+ .class_data = (void *) &bcm283x_socs[i],
120
+ };
121
+ type_register(&ti);
122
+ }
123
}
124
125
type_init(bcm2836_register_types)
126
diff --git a/hw/arm/raspi.c b/hw/arm/raspi.c
127
index XXXXXXX..XXXXXXX 100644
128
--- a/hw/arm/raspi.c
129
+++ b/hw/arm/raspi.c
130
@@ -XXX,XX +XXX,XX @@ static void raspi_init(MachineState *machine, int version)
131
BusState *bus;
132
DeviceState *carddev;
133
134
- object_initialize(&s->soc, sizeof(s->soc), TYPE_BCM283X);
135
+ object_initialize(&s->soc, sizeof(s->soc),
136
+ version == 3 ? TYPE_BCM2837 : TYPE_BCM2836);
137
object_property_add_child(OBJECT(machine), "soc", OBJECT(&s->soc),
138
&error_abort);
139
32
140
--
33
--
141
2.16.2
34
2.20.1
142
35
143
36
diff view generated by jsdifflib
1
From: Wei Huang <wei@redhat.com>
1
From: Andrew Jones <drjones@redhat.com>
2
2
3
For guest kernel that supports KASLR, the load address can change every
3
If we know what the default value should be then we can test for
4
time when guest VM runs. To find the physical base address correctly,
4
that as well as the feature existence.
5
current QEMU dump searches VMCOREINFO for the string "NUMBER(phys_base)=".
6
However this string pattern is only available on x86_64. AArch64 uses a
7
different field, called "NUMBER(PHYS_OFFSET)=". This patch makes sure
8
QEMU dump uses the correct string on AArch64.
9
5
10
Signed-off-by: Wei Huang <wei@redhat.com>
6
Signed-off-by: Andrew Jones <drjones@redhat.com>
11
Reviewed-by: Marc-André Lureau <marcandre.lureau@redhat.com>
7
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
12
Message-id: 1520615003-20869-1-git-send-email-wei@redhat.com
8
Message-id: 20200120101023.16030-5-drjones@redhat.com
13
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
9
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
14
---
10
---
15
dump.c | 14 +++++++++++---
11
tests/qtest/arm-cpu-features.c | 37 +++++++++++++++++++++++++---------
16
1 file changed, 11 insertions(+), 3 deletions(-)
12
1 file changed, 28 insertions(+), 9 deletions(-)
17
13
18
diff --git a/dump.c b/dump.c
14
diff --git a/tests/qtest/arm-cpu-features.c b/tests/qtest/arm-cpu-features.c
19
index XXXXXXX..XXXXXXX 100644
15
index XXXXXXX..XXXXXXX 100644
20
--- a/dump.c
16
--- a/tests/qtest/arm-cpu-features.c
21
+++ b/dump.c
17
+++ b/tests/qtest/arm-cpu-features.c
22
@@ -XXX,XX +XXX,XX @@ static void vmcoreinfo_update_phys_base(DumpState *s)
18
@@ -XXX,XX +XXX,XX @@ static bool resp_get_feature(QDict *resp, const char *feature)
23
19
qobject_unref(_resp); \
24
lines = g_strsplit((char *)vmci, "\n", -1);
20
})
25
for (i = 0; lines[i]; i++) {
21
26
- if (g_str_has_prefix(lines[i], "NUMBER(phys_base)=")) {
22
+#define assert_feature(qts, cpu_type, feature, expected_value) \
27
- if (qemu_strtou64(lines[i] + 18, NULL, 16,
23
+({ \
28
+ const char *prefix = NULL;
24
+ QDict *_resp, *_props; \
25
+ \
26
+ _resp = do_query_no_props(qts, cpu_type); \
27
+ g_assert(_resp); \
28
+ g_assert(resp_has_props(_resp)); \
29
+ _props = resp_get_props(_resp); \
30
+ g_assert(qdict_get(_props, feature)); \
31
+ g_assert(qdict_get_bool(_props, feature) == (expected_value)); \
32
+ qobject_unref(_resp); \
33
+})
29
+
34
+
30
+ if (s->dump_info.d_machine == EM_X86_64) {
35
+#define assert_has_feature_enabled(qts, cpu_type, feature) \
31
+ prefix = "NUMBER(phys_base)=";
36
+ assert_feature(qts, cpu_type, feature, true)
32
+ } else if (s->dump_info.d_machine == EM_AARCH64) {
33
+ prefix = "NUMBER(PHYS_OFFSET)=";
34
+ }
35
+
37
+
36
+ if (prefix && g_str_has_prefix(lines[i], prefix)) {
38
+#define assert_has_feature_disabled(qts, cpu_type, feature) \
37
+ if (qemu_strtou64(lines[i] + strlen(prefix), NULL, 16,
39
+ assert_feature(qts, cpu_type, feature, false)
38
&phys_base) < 0) {
40
+
39
- warn_report("Failed to read NUMBER(phys_base)=");
41
static void assert_type_full(QTestState *qts)
40
+ warn_report("Failed to read %s", prefix);
42
{
41
} else {
43
const char *error;
42
s->dump_info.phys_base = phys_base;
44
@@ -XXX,XX +XXX,XX @@ static void test_query_cpu_model_expansion(const void *data)
43
}
45
assert_error(qts, "host", "The CPU type 'host' requires KVM", NULL);
46
47
/* Test expected feature presence/absence for some cpu types */
48
- assert_has_feature(qts, "max", "pmu");
49
- assert_has_feature(qts, "cortex-a15", "pmu");
50
+ assert_has_feature_enabled(qts, "max", "pmu");
51
+ assert_has_feature_enabled(qts, "cortex-a15", "pmu");
52
assert_has_not_feature(qts, "cortex-a15", "aarch64");
53
54
if (g_str_equal(qtest_get_arch(), "aarch64")) {
55
- assert_has_feature(qts, "max", "aarch64");
56
- assert_has_feature(qts, "max", "sve");
57
- assert_has_feature(qts, "max", "sve128");
58
- assert_has_feature(qts, "cortex-a57", "pmu");
59
- assert_has_feature(qts, "cortex-a57", "aarch64");
60
+ assert_has_feature_enabled(qts, "max", "aarch64");
61
+ assert_has_feature_enabled(qts, "max", "sve");
62
+ assert_has_feature_enabled(qts, "max", "sve128");
63
+ assert_has_feature_enabled(qts, "cortex-a57", "pmu");
64
+ assert_has_feature_enabled(qts, "cortex-a57", "aarch64");
65
66
sve_tests_default(qts, "max");
67
68
@@ -XXX,XX +XXX,XX @@ static void test_query_cpu_model_expansion_kvm(const void *data)
69
QDict *resp;
70
char *error;
71
72
- assert_has_feature(qts, "host", "aarch64");
73
- assert_has_feature(qts, "host", "pmu");
74
+ assert_has_feature_enabled(qts, "host", "aarch64");
75
+ assert_has_feature_enabled(qts, "host", "pmu");
76
77
assert_error(qts, "cortex-a15",
78
"We cannot guarantee the CPU type 'cortex-a15' works "
44
--
79
--
45
2.16.2
80
2.20.1
46
81
47
82
diff view generated by jsdifflib
New patch
1
1
From: Andrew Jones <drjones@redhat.com>
2
3
When a VM is stopped (such as when it's paused) guest virtual time
4
should stop counting. Otherwise, when the VM is resumed it will
5
experience time jumps and its kernel may report soft lockups. Not
6
counting virtual time while the VM is stopped has the side effect
7
of making the guest's time appear to lag when compared with real
8
time, and even with time derived from the physical counter. For
9
this reason, this change, which is enabled by default, comes with
10
a KVM CPU feature allowing it to be disabled, restoring legacy
11
behavior.
12
13
This patch only provides the implementation of the virtual time
14
adjustment. A subsequent patch will provide the CPU property
15
allowing the change to be enabled and disabled.
16
17
Reported-by: Bijan Mottahedeh <bijan.mottahedeh@oracle.com>
18
Signed-off-by: Andrew Jones <drjones@redhat.com>
19
Message-id: 20200120101023.16030-6-drjones@redhat.com
20
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
21
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
22
---
23
target/arm/cpu.h | 7 ++++
24
target/arm/kvm_arm.h | 38 ++++++++++++++++++
25
target/arm/kvm.c | 92 ++++++++++++++++++++++++++++++++++++++++++++
26
target/arm/kvm32.c | 3 ++
27
target/arm/kvm64.c | 3 ++
28
target/arm/machine.c | 7 ++++
29
6 files changed, 150 insertions(+)
30
31
diff --git a/target/arm/cpu.h b/target/arm/cpu.h
32
index XXXXXXX..XXXXXXX 100644
33
--- a/target/arm/cpu.h
34
+++ b/target/arm/cpu.h
35
@@ -XXX,XX +XXX,XX @@ struct ARMCPU {
36
/* KVM init features for this CPU */
37
uint32_t kvm_init_features[7];
38
39
+ /* KVM CPU state */
40
+
41
+ /* KVM virtual time adjustment */
42
+ bool kvm_adjvtime;
43
+ bool kvm_vtime_dirty;
44
+ uint64_t kvm_vtime;
45
+
46
/* Uniprocessor system with MP extensions */
47
bool mp_is_up;
48
49
diff --git a/target/arm/kvm_arm.h b/target/arm/kvm_arm.h
50
index XXXXXXX..XXXXXXX 100644
51
--- a/target/arm/kvm_arm.h
52
+++ b/target/arm/kvm_arm.h
53
@@ -XXX,XX +XXX,XX @@ bool write_list_to_kvmstate(ARMCPU *cpu, int level);
54
*/
55
bool write_kvmstate_to_list(ARMCPU *cpu);
56
57
+/**
58
+ * kvm_arm_cpu_pre_save:
59
+ * @cpu: ARMCPU
60
+ *
61
+ * Called after write_kvmstate_to_list() from cpu_pre_save() to update
62
+ * the cpreg list with KVM CPU state.
63
+ */
64
+void kvm_arm_cpu_pre_save(ARMCPU *cpu);
65
+
66
+/**
67
+ * kvm_arm_cpu_post_load:
68
+ * @cpu: ARMCPU
69
+ *
70
+ * Called from cpu_post_load() to update KVM CPU state from the cpreg list.
71
+ */
72
+void kvm_arm_cpu_post_load(ARMCPU *cpu);
73
+
74
/**
75
* kvm_arm_reset_vcpu:
76
* @cpu: ARMCPU
77
@@ -XXX,XX +XXX,XX @@ int kvm_arm_sync_mpstate_to_kvm(ARMCPU *cpu);
78
*/
79
int kvm_arm_sync_mpstate_to_qemu(ARMCPU *cpu);
80
81
+/**
82
+ * kvm_arm_get_virtual_time:
83
+ * @cs: CPUState
84
+ *
85
+ * Gets the VCPU's virtual counter and stores it in the KVM CPU state.
86
+ */
87
+void kvm_arm_get_virtual_time(CPUState *cs);
88
+
89
+/**
90
+ * kvm_arm_put_virtual_time:
91
+ * @cs: CPUState
92
+ *
93
+ * Sets the VCPU's virtual counter to the value stored in the KVM CPU state.
94
+ */
95
+void kvm_arm_put_virtual_time(CPUState *cs);
96
+
97
+void kvm_arm_vm_state_change(void *opaque, int running, RunState state);
98
+
99
int kvm_arm_vgic_probe(void);
100
101
void kvm_arm_pmu_set_irq(CPUState *cs, int irq);
102
@@ -XXX,XX +XXX,XX @@ static inline void kvm_arm_pmu_set_irq(CPUState *cs, int irq) {}
103
static inline void kvm_arm_pmu_init(CPUState *cs) {}
104
105
static inline void kvm_arm_sve_get_vls(CPUState *cs, unsigned long *map) {}
106
+
107
+static inline void kvm_arm_get_virtual_time(CPUState *cs) {}
108
+static inline void kvm_arm_put_virtual_time(CPUState *cs) {}
109
#endif
110
111
static inline const char *gic_class_name(void)
112
diff --git a/target/arm/kvm.c b/target/arm/kvm.c
113
index XXXXXXX..XXXXXXX 100644
114
--- a/target/arm/kvm.c
115
+++ b/target/arm/kvm.c
116
@@ -XXX,XX +XXX,XX @@ static int compare_u64(const void *a, const void *b)
117
return 0;
118
}
119
120
+/*
121
+ * cpreg_values are sorted in ascending order by KVM register ID
122
+ * (see kvm_arm_init_cpreg_list). This allows us to cheaply find
123
+ * the storage for a KVM register by ID with a binary search.
124
+ */
125
+static uint64_t *kvm_arm_get_cpreg_ptr(ARMCPU *cpu, uint64_t regidx)
126
+{
127
+ uint64_t *res;
128
+
129
+ res = bsearch(&regidx, cpu->cpreg_indexes, cpu->cpreg_array_len,
130
+ sizeof(uint64_t), compare_u64);
131
+ assert(res);
132
+
133
+ return &cpu->cpreg_values[res - cpu->cpreg_indexes];
134
+}
135
+
136
/* Initialize the ARMCPU cpreg list according to the kernel's
137
* definition of what CPU registers it knows about (and throw away
138
* the previous TCG-created cpreg list).
139
@@ -XXX,XX +XXX,XX @@ bool write_list_to_kvmstate(ARMCPU *cpu, int level)
140
return ok;
141
}
142
143
+void kvm_arm_cpu_pre_save(ARMCPU *cpu)
144
+{
145
+ /* KVM virtual time adjustment */
146
+ if (cpu->kvm_vtime_dirty) {
147
+ *kvm_arm_get_cpreg_ptr(cpu, KVM_REG_ARM_TIMER_CNT) = cpu->kvm_vtime;
148
+ }
149
+}
150
+
151
+void kvm_arm_cpu_post_load(ARMCPU *cpu)
152
+{
153
+ /* KVM virtual time adjustment */
154
+ if (cpu->kvm_adjvtime) {
155
+ cpu->kvm_vtime = *kvm_arm_get_cpreg_ptr(cpu, KVM_REG_ARM_TIMER_CNT);
156
+ cpu->kvm_vtime_dirty = true;
157
+ }
158
+}
159
+
160
void kvm_arm_reset_vcpu(ARMCPU *cpu)
161
{
162
int ret;
163
@@ -XXX,XX +XXX,XX @@ int kvm_arm_sync_mpstate_to_qemu(ARMCPU *cpu)
164
return 0;
165
}
166
167
+void kvm_arm_get_virtual_time(CPUState *cs)
168
+{
169
+ ARMCPU *cpu = ARM_CPU(cs);
170
+ struct kvm_one_reg reg = {
171
+ .id = KVM_REG_ARM_TIMER_CNT,
172
+ .addr = (uintptr_t)&cpu->kvm_vtime,
173
+ };
174
+ int ret;
175
+
176
+ if (cpu->kvm_vtime_dirty) {
177
+ return;
178
+ }
179
+
180
+ ret = kvm_vcpu_ioctl(cs, KVM_GET_ONE_REG, &reg);
181
+ if (ret) {
182
+ error_report("Failed to get KVM_REG_ARM_TIMER_CNT");
183
+ abort();
184
+ }
185
+
186
+ cpu->kvm_vtime_dirty = true;
187
+}
188
+
189
+void kvm_arm_put_virtual_time(CPUState *cs)
190
+{
191
+ ARMCPU *cpu = ARM_CPU(cs);
192
+ struct kvm_one_reg reg = {
193
+ .id = KVM_REG_ARM_TIMER_CNT,
194
+ .addr = (uintptr_t)&cpu->kvm_vtime,
195
+ };
196
+ int ret;
197
+
198
+ if (!cpu->kvm_vtime_dirty) {
199
+ return;
200
+ }
201
+
202
+ ret = kvm_vcpu_ioctl(cs, KVM_SET_ONE_REG, &reg);
203
+ if (ret) {
204
+ error_report("Failed to set KVM_REG_ARM_TIMER_CNT");
205
+ abort();
206
+ }
207
+
208
+ cpu->kvm_vtime_dirty = false;
209
+}
210
+
211
int kvm_put_vcpu_events(ARMCPU *cpu)
212
{
213
CPUARMState *env = &cpu->env;
214
@@ -XXX,XX +XXX,XX @@ MemTxAttrs kvm_arch_post_run(CPUState *cs, struct kvm_run *run)
215
return MEMTXATTRS_UNSPECIFIED;
216
}
217
218
+void kvm_arm_vm_state_change(void *opaque, int running, RunState state)
219
+{
220
+ CPUState *cs = opaque;
221
+ ARMCPU *cpu = ARM_CPU(cs);
222
+
223
+ if (running) {
224
+ if (cpu->kvm_adjvtime) {
225
+ kvm_arm_put_virtual_time(cs);
226
+ }
227
+ } else {
228
+ if (cpu->kvm_adjvtime) {
229
+ kvm_arm_get_virtual_time(cs);
230
+ }
231
+ }
232
+}
233
234
int kvm_arch_handle_exit(CPUState *cs, struct kvm_run *run)
235
{
236
diff --git a/target/arm/kvm32.c b/target/arm/kvm32.c
237
index XXXXXXX..XXXXXXX 100644
238
--- a/target/arm/kvm32.c
239
+++ b/target/arm/kvm32.c
240
@@ -XXX,XX +XXX,XX @@
241
#include "qemu-common.h"
242
#include "cpu.h"
243
#include "qemu/timer.h"
244
+#include "sysemu/runstate.h"
245
#include "sysemu/kvm.h"
246
#include "kvm_arm.h"
247
#include "internals.h"
248
@@ -XXX,XX +XXX,XX @@ int kvm_arch_init_vcpu(CPUState *cs)
249
return -EINVAL;
250
}
251
252
+ qemu_add_vm_change_state_handler(kvm_arm_vm_state_change, cs);
253
+
254
/* Determine init features for this CPU */
255
memset(cpu->kvm_init_features, 0, sizeof(cpu->kvm_init_features));
256
if (cpu->start_powered_off) {
257
diff --git a/target/arm/kvm64.c b/target/arm/kvm64.c
258
index XXXXXXX..XXXXXXX 100644
259
--- a/target/arm/kvm64.c
260
+++ b/target/arm/kvm64.c
261
@@ -XXX,XX +XXX,XX @@
262
#include "qemu/host-utils.h"
263
#include "qemu/main-loop.h"
264
#include "exec/gdbstub.h"
265
+#include "sysemu/runstate.h"
266
#include "sysemu/kvm.h"
267
#include "sysemu/kvm_int.h"
268
#include "kvm_arm.h"
269
@@ -XXX,XX +XXX,XX @@ int kvm_arch_init_vcpu(CPUState *cs)
270
return -EINVAL;
271
}
272
273
+ qemu_add_vm_change_state_handler(kvm_arm_vm_state_change, cs);
274
+
275
/* Determine init features for this CPU */
276
memset(cpu->kvm_init_features, 0, sizeof(cpu->kvm_init_features));
277
if (cpu->start_powered_off) {
278
diff --git a/target/arm/machine.c b/target/arm/machine.c
279
index XXXXXXX..XXXXXXX 100644
280
--- a/target/arm/machine.c
281
+++ b/target/arm/machine.c
282
@@ -XXX,XX +XXX,XX @@ static int cpu_pre_save(void *opaque)
283
/* This should never fail */
284
abort();
285
}
286
+
287
+ /*
288
+ * kvm_arm_cpu_pre_save() must be called after
289
+ * write_kvmstate_to_list()
290
+ */
291
+ kvm_arm_cpu_pre_save(cpu);
292
} else {
293
if (!write_cpustate_to_list(cpu, false)) {
294
/* This should never fail. */
295
@@ -XXX,XX +XXX,XX @@ static int cpu_post_load(void *opaque, int version_id)
296
* we're using it.
297
*/
298
write_list_to_cpustate(cpu);
299
+ kvm_arm_cpu_post_load(cpu);
300
} else {
301
if (!write_list_to_cpustate(cpu)) {
302
return -1;
303
--
304
2.20.1
305
306
diff view generated by jsdifflib
New patch
1
1
From: Andrew Jones <drjones@redhat.com>
2
3
kvm-no-adjvtime is a KVM specific CPU property and a first of its
4
kind. To accommodate it we also add kvm_arm_add_vcpu_properties()
5
and a KVM specific CPU properties description to the CPU features
6
document.
7
8
Signed-off-by: Andrew Jones <drjones@redhat.com>
9
Message-id: 20200120101023.16030-7-drjones@redhat.com
10
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
11
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
12
---
13
include/hw/arm/virt.h | 1 +
14
target/arm/kvm_arm.h | 11 ++++++++++
15
hw/arm/virt.c | 8 ++++++++
16
target/arm/cpu.c | 2 ++
17
target/arm/cpu64.c | 1 +
18
target/arm/kvm.c | 28 +++++++++++++++++++++++++
19
target/arm/monitor.c | 1 +
20
tests/qtest/arm-cpu-features.c | 4 ++++
21
docs/arm-cpu-features.rst | 37 +++++++++++++++++++++++++++++++++-
22
9 files changed, 92 insertions(+), 1 deletion(-)
23
24
diff --git a/include/hw/arm/virt.h b/include/hw/arm/virt.h
25
index XXXXXXX..XXXXXXX 100644
26
--- a/include/hw/arm/virt.h
27
+++ b/include/hw/arm/virt.h
28
@@ -XXX,XX +XXX,XX @@ typedef struct {
29
bool smbios_old_sys_ver;
30
bool no_highmem_ecam;
31
bool no_ged; /* Machines < 4.2 has no support for ACPI GED device */
32
+ bool kvm_no_adjvtime;
33
} VirtMachineClass;
34
35
typedef struct {
36
diff --git a/target/arm/kvm_arm.h b/target/arm/kvm_arm.h
37
index XXXXXXX..XXXXXXX 100644
38
--- a/target/arm/kvm_arm.h
39
+++ b/target/arm/kvm_arm.h
40
@@ -XXX,XX +XXX,XX @@ void kvm_arm_sve_get_vls(CPUState *cs, unsigned long *map);
41
*/
42
void kvm_arm_set_cpu_features_from_host(ARMCPU *cpu);
43
44
+/**
45
+ * kvm_arm_add_vcpu_properties:
46
+ * @obj: The CPU object to add the properties to
47
+ *
48
+ * Add all KVM specific CPU properties to the CPU object. These
49
+ * are the CPU properties with "kvm-" prefixed names.
50
+ */
51
+void kvm_arm_add_vcpu_properties(Object *obj);
52
+
53
/**
54
* kvm_arm_aarch32_supported:
55
* @cs: CPUState
56
@@ -XXX,XX +XXX,XX @@ static inline void kvm_arm_set_cpu_features_from_host(ARMCPU *cpu)
57
cpu->host_cpu_probe_failed = true;
58
}
59
60
+static inline void kvm_arm_add_vcpu_properties(Object *obj) {}
61
+
62
static inline bool kvm_arm_aarch32_supported(CPUState *cs)
63
{
64
return false;
65
diff --git a/hw/arm/virt.c b/hw/arm/virt.c
66
index XXXXXXX..XXXXXXX 100644
67
--- a/hw/arm/virt.c
68
+++ b/hw/arm/virt.c
69
@@ -XXX,XX +XXX,XX @@ static void machvirt_init(MachineState *machine)
70
}
71
}
72
73
+ if (vmc->kvm_no_adjvtime &&
74
+ object_property_find(cpuobj, "kvm-no-adjvtime", NULL)) {
75
+ object_property_set_bool(cpuobj, true, "kvm-no-adjvtime", NULL);
76
+ }
77
+
78
if (vmc->no_pmu && object_property_find(cpuobj, "pmu", NULL)) {
79
object_property_set_bool(cpuobj, false, "pmu", NULL);
80
}
81
@@ -XXX,XX +XXX,XX @@ DEFINE_VIRT_MACHINE_AS_LATEST(5, 0)
82
83
static void virt_machine_4_2_options(MachineClass *mc)
84
{
85
+ VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
86
+
87
virt_machine_5_0_options(mc);
88
compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
89
+ vmc->kvm_no_adjvtime = true;
90
}
91
DEFINE_VIRT_MACHINE(4, 2)
92
93
diff --git a/target/arm/cpu.c b/target/arm/cpu.c
94
index XXXXXXX..XXXXXXX 100644
95
--- a/target/arm/cpu.c
96
+++ b/target/arm/cpu.c
97
@@ -XXX,XX +XXX,XX @@ static void arm_max_initfn(Object *obj)
98
99
if (kvm_enabled()) {
100
kvm_arm_set_cpu_features_from_host(cpu);
101
+ kvm_arm_add_vcpu_properties(obj);
102
} else {
103
cortex_a15_initfn(obj);
104
105
@@ -XXX,XX +XXX,XX @@ static void arm_host_initfn(Object *obj)
106
if (arm_feature(&cpu->env, ARM_FEATURE_AARCH64)) {
107
aarch64_add_sve_properties(obj);
108
}
109
+ kvm_arm_add_vcpu_properties(obj);
110
arm_cpu_post_init(obj);
111
}
112
113
diff --git a/target/arm/cpu64.c b/target/arm/cpu64.c
114
index XXXXXXX..XXXXXXX 100644
115
--- a/target/arm/cpu64.c
116
+++ b/target/arm/cpu64.c
117
@@ -XXX,XX +XXX,XX @@ static void aarch64_max_initfn(Object *obj)
118
119
if (kvm_enabled()) {
120
kvm_arm_set_cpu_features_from_host(cpu);
121
+ kvm_arm_add_vcpu_properties(obj);
122
} else {
123
uint64_t t;
124
uint32_t u;
125
diff --git a/target/arm/kvm.c b/target/arm/kvm.c
126
index XXXXXXX..XXXXXXX 100644
127
--- a/target/arm/kvm.c
128
+++ b/target/arm/kvm.c
129
@@ -XXX,XX +XXX,XX @@
130
#include "qemu/timer.h"
131
#include "qemu/error-report.h"
132
#include "qemu/main-loop.h"
133
+#include "qom/object.h"
134
+#include "qapi/error.h"
135
#include "sysemu/sysemu.h"
136
#include "sysemu/kvm.h"
137
#include "sysemu/kvm_int.h"
138
@@ -XXX,XX +XXX,XX @@ void kvm_arm_set_cpu_features_from_host(ARMCPU *cpu)
139
env->features = arm_host_cpu_features.features;
140
}
141
142
+static bool kvm_no_adjvtime_get(Object *obj, Error **errp)
143
+{
144
+ return !ARM_CPU(obj)->kvm_adjvtime;
145
+}
146
+
147
+static void kvm_no_adjvtime_set(Object *obj, bool value, Error **errp)
148
+{
149
+ ARM_CPU(obj)->kvm_adjvtime = !value;
150
+}
151
+
152
+/* KVM VCPU properties should be prefixed with "kvm-". */
153
+void kvm_arm_add_vcpu_properties(Object *obj)
154
+{
155
+ if (!kvm_enabled()) {
156
+ return;
157
+ }
158
+
159
+ ARM_CPU(obj)->kvm_adjvtime = true;
160
+ object_property_add_bool(obj, "kvm-no-adjvtime", kvm_no_adjvtime_get,
161
+ kvm_no_adjvtime_set, &error_abort);
162
+ object_property_set_description(obj, "kvm-no-adjvtime",
163
+ "Set on to disable the adjustment of "
164
+ "the virtual counter. VM stopped time "
165
+ "will be counted.", &error_abort);
166
+}
167
+
168
bool kvm_arm_pmu_supported(CPUState *cpu)
169
{
170
return kvm_check_extension(cpu->kvm_state, KVM_CAP_ARM_PMU_V3);
171
diff --git a/target/arm/monitor.c b/target/arm/monitor.c
172
index XXXXXXX..XXXXXXX 100644
173
--- a/target/arm/monitor.c
174
+++ b/target/arm/monitor.c
175
@@ -XXX,XX +XXX,XX @@ static const char *cpu_model_advertised_features[] = {
176
"sve128", "sve256", "sve384", "sve512",
177
"sve640", "sve768", "sve896", "sve1024", "sve1152", "sve1280",
178
"sve1408", "sve1536", "sve1664", "sve1792", "sve1920", "sve2048",
179
+ "kvm-no-adjvtime",
180
NULL
181
};
182
183
diff --git a/tests/qtest/arm-cpu-features.c b/tests/qtest/arm-cpu-features.c
184
index XXXXXXX..XXXXXXX 100644
185
--- a/tests/qtest/arm-cpu-features.c
186
+++ b/tests/qtest/arm-cpu-features.c
187
@@ -XXX,XX +XXX,XX @@ static void test_query_cpu_model_expansion(const void *data)
188
assert_has_feature_enabled(qts, "cortex-a15", "pmu");
189
assert_has_not_feature(qts, "cortex-a15", "aarch64");
190
191
+ assert_has_not_feature(qts, "max", "kvm-no-adjvtime");
192
+
193
if (g_str_equal(qtest_get_arch(), "aarch64")) {
194
assert_has_feature_enabled(qts, "max", "aarch64");
195
assert_has_feature_enabled(qts, "max", "sve");
196
@@ -XXX,XX +XXX,XX @@ static void test_query_cpu_model_expansion_kvm(const void *data)
197
return;
198
}
199
200
+ assert_has_feature_disabled(qts, "host", "kvm-no-adjvtime");
201
+
202
if (g_str_equal(qtest_get_arch(), "aarch64")) {
203
bool kvm_supports_sve;
204
char max_name[8], name[8];
205
diff --git a/docs/arm-cpu-features.rst b/docs/arm-cpu-features.rst
206
index XXXXXXX..XXXXXXX 100644
207
--- a/docs/arm-cpu-features.rst
208
+++ b/docs/arm-cpu-features.rst
209
@@ -XXX,XX +XXX,XX @@ supporting the feature or only supporting the feature under certain
210
configurations. For example, the `aarch64` CPU feature, which, when
211
disabled, enables the optional AArch32 CPU feature, is only supported
212
when using the KVM accelerator and when running on a host CPU type that
213
-supports the feature.
214
+supports the feature. While `aarch64` currently only works with KVM,
215
+it could work with TCG. CPU features that are specific to KVM are
216
+prefixed with "kvm-" and are described in "KVM VCPU Features".
217
218
CPU Feature Probing
219
===================
220
@@ -XXX,XX +XXX,XX @@ disabling many SVE vector lengths would be quite verbose, the `sve<N>` CPU
221
properties have special semantics (see "SVE CPU Property Parsing
222
Semantics").
223
224
+KVM VCPU Features
225
+=================
226
+
227
+KVM VCPU features are CPU features that are specific to KVM, such as
228
+paravirt features or features that enable CPU virtualization extensions.
229
+The features' CPU properties are only available when KVM is enabled and
230
+are named with the prefix "kvm-". KVM VCPU features may be probed,
231
+enabled, and disabled in the same way as other CPU features. Below is
232
+the list of KVM VCPU features and their descriptions.
233
+
234
+ kvm-no-adjvtime By default kvm-no-adjvtime is disabled. This
235
+ means that by default the virtual time
236
+ adjustment is enabled (vtime is *not not*
237
+ adjusted).
238
+
239
+ When virtual time adjustment is enabled each
240
+ time the VM transitions back to running state
241
+ the VCPU's virtual counter is updated to ensure
242
+ stopped time is not counted. This avoids time
243
+ jumps surprising guest OSes and applications,
244
+ as long as they use the virtual counter for
245
+ timekeeping. However it has the side effect of
246
+ the virtual and physical counters diverging.
247
+ All timekeeping based on the virtual counter
248
+ will appear to lag behind any timekeeping that
249
+ does not subtract VM stopped time. The guest
250
+ may resynchronize its virtual counter with
251
+ other time sources as needed.
252
+
253
+ Enable kvm-no-adjvtime to disable virtual time
254
+ adjustment, also restoring the legacy (pre-5.0)
255
+ behavior.
256
+
257
SVE CPU Properties
258
==================
259
260
--
261
2.20.1
262
263
diff view generated by jsdifflib