Patchwork pass info about hpets to seabios.

login
register
mail settings
Submitter Gleb Natapov
Date June 13, 2010, 2:43 p.m.
Message ID <20100613144315.GB6292@redhat.com>
Download mbox | patch
Permalink /patch/55442/
State New
Headers show

Comments

Gleb Natapov - June 13, 2010, 2:43 p.m.
Currently HPET ACPI table is created regardless of whether qemu actually
created hpet device. This may confuse some guests that don't check that
hpet is functional before using it. Solve this by passing info about
hpets in qemu to seabios via fw config interface. Additional benefit is
that seabios no longer uses hard coded hpet configuration. Proposed
interface supports up to 256 hpets. This is the number defined by hpet 
spec.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
--
			Gleb.
Jan Kiszka - June 13, 2010, 4:56 p.m.
Gleb Natapov wrote:
> Currently HPET ACPI table is created regardless of whether qemu actually
> created hpet device. This may confuse some guests that don't check that
> hpet is functional before using it. Solve this by passing info about
> hpets in qemu to seabios via fw config interface. Additional benefit is
> that seabios no longer uses hard coded hpet configuration. Proposed
> interface supports up to 256 hpets. This is the number defined by hpet 
> spec.

Nice, this lays the ground for adding hpets via -device.

(But I think I read there can only be 8 hpets with a total sum of 256
timers.)

> 
> Signed-off-by: Gleb Natapov <gleb@redhat.com>
> diff --git a/hw/hpet.c b/hw/hpet.c
> index 93fc399..f2a4514 100644
> --- a/hw/hpet.c
> +++ b/hw/hpet.c
> @@ -73,6 +73,8 @@ typedef struct HPETState {
>      uint64_t hpet_counter;      /* main counter */
>  } HPETState;
>  
> +struct hpet_fw_config hpet_cfg = {.valid = 1};
> +
>  static uint32_t hpet_in_legacy_mode(HPETState *s)
>  {
>      return s->config & HPET_CFG_LEGACY;
> @@ -661,6 +663,9 @@ static void hpet_reset(DeviceState *d)
>           */
>          hpet_pit_enable();
>      }
> +    hpet_cfg.count = 1;
> +    hpet_cfg.hpet.event_timer_block_id = (uint32_t)s->capability;

The number of timers, thus the content of capability can change on
vmload. So you need to update hpet_cfg there as well.

And I think we can move the capability setup into init. But this is not
directly related to this patch, would just avoid adding this hunk to
hpet_reset.

> +    hpet_cfg.hpet.address = sysbus_from_qdev(d)->mmio[0].addr;
>      count = 1;
>  }
>  
> diff --git a/hw/hpet_emul.h b/hw/hpet_emul.h
> index d7bc102..5cf5463 100644
> --- a/hw/hpet_emul.h
> +++ b/hw/hpet_emul.h
> @@ -53,4 +53,20 @@
>  #define HPET_TN_INT_ROUTE_CAP_SHIFT 32
>  #define HPET_TN_CFG_BITS_READONLY_OR_RESERVED 0xffff80b1U
>  
> +struct hpet_fw_entry
> +{
> +    uint32_t event_timer_block_id;
> +    uint64_t address;
> +    uint16_t min_tick;
> +    uint8_t page_prot;
> +} __attribute__ ((packed));
> +
> +struct hpet_fw_config
> +{
> +    uint8_t valid;
> +    uint8_t count;
> +    struct hpet_fw_entry hpet;

Why not already struct hpet_fw_entry hpet[8]? Once the bios bits are
merge, we can quickly remove the single hpet limitation on qemu side.

> +} __attribute__ ((packed));
> +
> +extern struct hpet_fw_config hpet_cfg;
>  #endif
> diff --git a/hw/pc.c b/hw/pc.c
> index 1491129..d14d657 100644
> --- a/hw/pc.c
> +++ b/hw/pc.c
> @@ -61,6 +61,7 @@
>  #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
>  #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
>  #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
> +#define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
>  
>  #define E820_NR_ENTRIES		16
>  
> @@ -484,6 +485,8 @@ static void *bochs_bios_init(void)
>      fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE, (uint8_t *)&e820_table,
>                       sizeof(struct e820_table));
>  
> +    fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, (uint8_t *)&hpet_cfg,
> +                     sizeof(struct hpet_fw_config));
>      /* allocate memory for the NUMA channel: one (64bit) word for the number
>       * of nodes, one word for each VCPU->node and one word for each node to
>       * hold the amount of memory.
> --
> 			Gleb.
> 
> 

Jan
Gleb Natapov - June 13, 2010, 5:19 p.m.
On Sun, Jun 13, 2010 at 06:56:37PM +0200, Jan Kiszka wrote:
> Gleb Natapov wrote:
> > Currently HPET ACPI table is created regardless of whether qemu actually
> > created hpet device. This may confuse some guests that don't check that
> > hpet is functional before using it. Solve this by passing info about
> > hpets in qemu to seabios via fw config interface. Additional benefit is
> > that seabios no longer uses hard coded hpet configuration. Proposed
> > interface supports up to 256 hpets. This is the number defined by hpet 
> > spec.
> 
> Nice, this lays the ground for adding hpets via -device.
> 
> (But I think I read there can only be 8 hpets with a total sum of 256
> timers.)
> 
Ah, correct. I thought to myself 256 hpets should be to much :)

> > 
> > Signed-off-by: Gleb Natapov <gleb@redhat.com>
> > diff --git a/hw/hpet.c b/hw/hpet.c
> > index 93fc399..f2a4514 100644
> > --- a/hw/hpet.c
> > +++ b/hw/hpet.c
> > @@ -73,6 +73,8 @@ typedef struct HPETState {
> >      uint64_t hpet_counter;      /* main counter */
> >  } HPETState;
> >  
> > +struct hpet_fw_config hpet_cfg = {.valid = 1};
> > +
> >  static uint32_t hpet_in_legacy_mode(HPETState *s)
> >  {
> >      return s->config & HPET_CFG_LEGACY;
> > @@ -661,6 +663,9 @@ static void hpet_reset(DeviceState *d)
> >           */
> >          hpet_pit_enable();
> >      }
> > +    hpet_cfg.count = 1;
> > +    hpet_cfg.hpet.event_timer_block_id = (uint32_t)s->capability;
> 
> The number of timers, thus the content of capability can change on
> vmload. So you need to update hpet_cfg there as well.
> 
How it can change? User is required to run the same command line on src
and dst, no?

> And I think we can move the capability setup into init. But this is not
> directly related to this patch, would just avoid adding this hunk to
> hpet_reset.
I actually did that initially and tried to init hpet_cfg there too, but
then noticed that mmio[0].addr below is not initialized at init time yet.

> 
> > +    hpet_cfg.hpet.address = sysbus_from_qdev(d)->mmio[0].addr;
> >      count = 1;
> >  }
> >  
> > diff --git a/hw/hpet_emul.h b/hw/hpet_emul.h
> > index d7bc102..5cf5463 100644
> > --- a/hw/hpet_emul.h
> > +++ b/hw/hpet_emul.h
> > @@ -53,4 +53,20 @@
> >  #define HPET_TN_INT_ROUTE_CAP_SHIFT 32
> >  #define HPET_TN_CFG_BITS_READONLY_OR_RESERVED 0xffff80b1U
> >  
> > +struct hpet_fw_entry
> > +{
> > +    uint32_t event_timer_block_id;
> > +    uint64_t address;
> > +    uint16_t min_tick;
> > +    uint8_t page_prot;
> > +} __attribute__ ((packed));
> > +
> > +struct hpet_fw_config
> > +{
> > +    uint8_t valid;
> > +    uint8_t count;
> > +    struct hpet_fw_entry hpet;
> 
> Why not already struct hpet_fw_entry hpet[8]? Once the bios bits are
> merge, we can quickly remove the single hpet limitation on qemu side.
> 
Number 256 somehow stuck in my head. 8 hpets is Ok to do from the start.

> > +} __attribute__ ((packed));
> > +
> > +extern struct hpet_fw_config hpet_cfg;
> >  #endif
> > diff --git a/hw/pc.c b/hw/pc.c
> > index 1491129..d14d657 100644
> > --- a/hw/pc.c
> > +++ b/hw/pc.c
> > @@ -61,6 +61,7 @@
> >  #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
> >  #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
> >  #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
> > +#define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
> >  
> >  #define E820_NR_ENTRIES		16
> >  
> > @@ -484,6 +485,8 @@ static void *bochs_bios_init(void)
> >      fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE, (uint8_t *)&e820_table,
> >                       sizeof(struct e820_table));
> >  
> > +    fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, (uint8_t *)&hpet_cfg,
> > +                     sizeof(struct hpet_fw_config));
> >      /* allocate memory for the NUMA channel: one (64bit) word for the number
> >       * of nodes, one word for each VCPU->node and one word for each node to
> >       * hold the amount of memory.
> > --
> > 			Gleb.
> > 
> > 
> 
> Jan
> 



--
			Gleb.
Jan Kiszka - June 13, 2010, 5:41 p.m.
Gleb Natapov wrote:
> On Sun, Jun 13, 2010 at 06:56:37PM +0200, Jan Kiszka wrote:
>> Gleb Natapov wrote:
>>> Currently HPET ACPI table is created regardless of whether qemu actually
>>> created hpet device. This may confuse some guests that don't check that
>>> hpet is functional before using it. Solve this by passing info about
>>> hpets in qemu to seabios via fw config interface. Additional benefit is
>>> that seabios no longer uses hard coded hpet configuration. Proposed
>>> interface supports up to 256 hpets. This is the number defined by hpet 
>>> spec.
>> Nice, this lays the ground for adding hpets via -device.
>>
>> (But I think I read there can only be 8 hpets with a total sum of 256
>> timers.)
>>
> Ah, correct. I thought to myself 256 hpets should be to much :)
> 
>>> Signed-off-by: Gleb Natapov <gleb@redhat.com>
>>> diff --git a/hw/hpet.c b/hw/hpet.c
>>> index 93fc399..f2a4514 100644
>>> --- a/hw/hpet.c
>>> +++ b/hw/hpet.c
>>> @@ -73,6 +73,8 @@ typedef struct HPETState {
>>>      uint64_t hpet_counter;      /* main counter */
>>>  } HPETState;
>>>  
>>> +struct hpet_fw_config hpet_cfg = {.valid = 1};
>>> +
>>>  static uint32_t hpet_in_legacy_mode(HPETState *s)
>>>  {
>>>      return s->config & HPET_CFG_LEGACY;
>>> @@ -661,6 +663,9 @@ static void hpet_reset(DeviceState *d)
>>>           */
>>>          hpet_pit_enable();
>>>      }
>>> +    hpet_cfg.count = 1;
>>> +    hpet_cfg.hpet.event_timer_block_id = (uint32_t)s->capability;
>> The number of timers, thus the content of capability can change on
>> vmload. So you need to update hpet_cfg there as well.
>>
> How it can change? User is required to run the same command line on src
> and dst, no?

Generally, yes. But there is no technical need to match the hpet props
so far, they are included in the vmstate (implicitly).

> 
>> And I think we can move the capability setup into init. But this is not
>> directly related to this patch, would just avoid adding this hunk to
>> hpet_reset.
> I actually did that initially and tried to init hpet_cfg there too, but
> then noticed that mmio[0].addr below is not initialized at init time yet.

Indeed.

You may try sysbus_init_mmio_cb instead for a callback to be invoked
once the address was set. That handler could also be called on vmload.

Jan
Gleb Natapov - June 13, 2010, 7:02 p.m.
On Sun, Jun 13, 2010 at 07:41:14PM +0200, Jan Kiszka wrote:
> Gleb Natapov wrote:
> > On Sun, Jun 13, 2010 at 06:56:37PM +0200, Jan Kiszka wrote:
> >> Gleb Natapov wrote:
> >>> Currently HPET ACPI table is created regardless of whether qemu actually
> >>> created hpet device. This may confuse some guests that don't check that
> >>> hpet is functional before using it. Solve this by passing info about
> >>> hpets in qemu to seabios via fw config interface. Additional benefit is
> >>> that seabios no longer uses hard coded hpet configuration. Proposed
> >>> interface supports up to 256 hpets. This is the number defined by hpet 
> >>> spec.
> >> Nice, this lays the ground for adding hpets via -device.
> >>
> >> (But I think I read there can only be 8 hpets with a total sum of 256
> >> timers.)
> >>
> > Ah, correct. I thought to myself 256 hpets should be to much :)
> > 
> >>> Signed-off-by: Gleb Natapov <gleb@redhat.com>
> >>> diff --git a/hw/hpet.c b/hw/hpet.c
> >>> index 93fc399..f2a4514 100644
> >>> --- a/hw/hpet.c
> >>> +++ b/hw/hpet.c
> >>> @@ -73,6 +73,8 @@ typedef struct HPETState {
> >>>      uint64_t hpet_counter;      /* main counter */
> >>>  } HPETState;
> >>>  
> >>> +struct hpet_fw_config hpet_cfg = {.valid = 1};
> >>> +
> >>>  static uint32_t hpet_in_legacy_mode(HPETState *s)
> >>>  {
> >>>      return s->config & HPET_CFG_LEGACY;
> >>> @@ -661,6 +663,9 @@ static void hpet_reset(DeviceState *d)
> >>>           */
> >>>          hpet_pit_enable();
> >>>      }
> >>> +    hpet_cfg.count = 1;
> >>> +    hpet_cfg.hpet.event_timer_block_id = (uint32_t)s->capability;
> >> The number of timers, thus the content of capability can change on
> >> vmload. So you need to update hpet_cfg there as well.
> >>
> > How it can change? User is required to run the same command line on src
> > and dst, no?
> 
> Generally, yes. But there is no technical need to match the hpet props
> so far, they are included in the vmstate (implicitly).
> 
Assumption that config at src and dst are identical is everywhere in the
code and in the future we should be able to pass actual config as part
of the migration protocol. Better fail migration if we can detect that
configs are different then trying to fix some things, but not others.

> > 
> >> And I think we can move the capability setup into init. But this is not
> >> directly related to this patch, would just avoid adding this hunk to
> >> hpet_reset.
> > I actually did that initially and tried to init hpet_cfg there too, but
> > then noticed that mmio[0].addr below is not initialized at init time yet.
> 
> Indeed.
> 
> You may try sysbus_init_mmio_cb instead for a callback to be invoked
> once the address was set. That handler could also be called on vmload.
> 
sysbus_init_mmio_cb calls callback instead of
cpu_register_physical_memory(). Looks like unneeded complication. Doing
it at reset should be fine as far as I see.

--
			Gleb.
Paul Brook - June 13, 2010, 7:55 p.m.
> > And I think we can move the capability setup into init. But this is not
> > directly related to this patch, would just avoid adding this hunk to
> > hpet_reset.
> 
> I actually did that initially and tried to init hpet_cfg there too, but
> then noticed that mmio[0].addr below is not initialized at init time yet.
> 
> > > +    hpet_cfg.hpet.address = sysbus_from_qdev(d)->mmio[0].addr;

This is one of the reasons that you're not supposed to be messing with 
DeviceState contents directly.

Paul
Gleb Natapov - June 14, 2010, 5:18 a.m.
On Sun, Jun 13, 2010 at 08:55:23PM +0100, Paul Brook wrote:
> > > And I think we can move the capability setup into init. But this is not
> > > directly related to this patch, would just avoid adding this hunk to
> > > hpet_reset.
> > 
> > I actually did that initially and tried to init hpet_cfg there too, but
> > then noticed that mmio[0].addr below is not initialized at init time yet.
> > 
> > > > +    hpet_cfg.hpet.address = sysbus_from_qdev(d)->mmio[0].addr;
> 
> This is one of the reasons that you're not supposed to be messing with 
> DeviceState contents directly.
> 
Alternatives?

--
			Gleb.

Patch

diff --git a/hw/hpet.c b/hw/hpet.c
index 93fc399..f2a4514 100644
--- a/hw/hpet.c
+++ b/hw/hpet.c
@@ -73,6 +73,8 @@  typedef struct HPETState {
     uint64_t hpet_counter;      /* main counter */
 } HPETState;
 
+struct hpet_fw_config hpet_cfg = {.valid = 1};
+
 static uint32_t hpet_in_legacy_mode(HPETState *s)
 {
     return s->config & HPET_CFG_LEGACY;
@@ -661,6 +663,9 @@  static void hpet_reset(DeviceState *d)
          */
         hpet_pit_enable();
     }
+    hpet_cfg.count = 1;
+    hpet_cfg.hpet.event_timer_block_id = (uint32_t)s->capability;
+    hpet_cfg.hpet.address = sysbus_from_qdev(d)->mmio[0].addr;
     count = 1;
 }
 
diff --git a/hw/hpet_emul.h b/hw/hpet_emul.h
index d7bc102..5cf5463 100644
--- a/hw/hpet_emul.h
+++ b/hw/hpet_emul.h
@@ -53,4 +53,20 @@ 
 #define HPET_TN_INT_ROUTE_CAP_SHIFT 32
 #define HPET_TN_CFG_BITS_READONLY_OR_RESERVED 0xffff80b1U
 
+struct hpet_fw_entry
+{
+    uint32_t event_timer_block_id;
+    uint64_t address;
+    uint16_t min_tick;
+    uint8_t page_prot;
+} __attribute__ ((packed));
+
+struct hpet_fw_config
+{
+    uint8_t valid;
+    uint8_t count;
+    struct hpet_fw_entry hpet;
+} __attribute__ ((packed));
+
+extern struct hpet_fw_config hpet_cfg;
 #endif
diff --git a/hw/pc.c b/hw/pc.c
index 1491129..d14d657 100644
--- a/hw/pc.c
+++ b/hw/pc.c
@@ -61,6 +61,7 @@ 
 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
+#define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
 
 #define E820_NR_ENTRIES		16
 
@@ -484,6 +485,8 @@  static void *bochs_bios_init(void)
     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE, (uint8_t *)&e820_table,
                      sizeof(struct e820_table));
 
+    fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, (uint8_t *)&hpet_cfg,
+                     sizeof(struct hpet_fw_config));
     /* allocate memory for the NUMA channel: one (64bit) word for the number
      * of nodes, one word for each VCPU->node and one word for each node to
      * hold the amount of memory.