[Qemu-devel] [PATCH 1/2] ioport: add function to check whenever a port is assigned or not

[Qemu-devel] [PATCH 1/2] ioport: add function to check whenever a port is assigned or not

[Gerd Hoffmann]

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
---
 ioport.c |    5 +++++
 ioport.h |    1 +
 2 files changed, 6 insertions(+), 0 deletions(-)

diff --git a/ioport.c b/ioport.c
index 53dd87a..b718047 100644
--- a/ioport.c
+++ b/ioport.c
@@ -190,6 +190,11 @@ void isa_unassign_ioport(pio_addr_t start, int length)
     }
 }
 
+int is_ioport_assigned(pio_addr_t addr)
+{
+    return ioport_opaque[addr] != NULL;
+}
+
 /***********************************************************/
 
 void cpu_outb(pio_addr_t addr, uint8_t val)
diff --git a/ioport.h b/ioport.h
index 3d3c8a3..46fbfa8 100644
--- a/ioport.h
+++ b/ioport.h
@@ -41,6 +41,7 @@ int register_ioport_read(pio_addr_t start, int length, int size,
 int register_ioport_write(pio_addr_t start, int length, int size,
                           IOPortWriteFunc *func, void *opaque);
 void isa_unassign_ioport(pio_addr_t start, int length);
+int is_ioport_assigned(pio_addr_t addr);
 
 
 void cpu_outb(pio_addr_t addr, uint8_t val);
-- 
1.6.6.1

[Qemu-devel] [PATCH 2/2] all vga: fail graicefully when vga ports are taken

[Gerd Hoffmann]

Try to pci hotplug a vga card, watch qemu die with hw_error().
This patch fixes it.

[v2: code style fixes]

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
---
 hw/cirrus_vga.c |    4 ++++
 hw/vga-pci.c    |    4 ++++
 hw/vmware_vga.c |    4 ++++
 3 files changed, 12 insertions(+), 0 deletions(-)

diff --git a/hw/cirrus_vga.c b/hw/cirrus_vga.c
index 9f61a01..977abd2 100644
--- a/hw/cirrus_vga.c
+++ b/hw/cirrus_vga.c
@@ -3189,6 +3189,10 @@ static int pci_cirrus_vga_initfn(PCIDevice *dev)
      uint8_t *pci_conf = d->dev.config;
      int device_id = CIRRUS_ID_CLGD5446;
 
+     if (is_ioport_assigned(0x3c0)) {
+         return -1;
+     }
+
      /* setup VGA */
      vga_common_init(&s->vga, VGA_RAM_SIZE);
      cirrus_init_common(s, device_id, 1);
diff --git a/hw/vga-pci.c b/hw/vga-pci.c
index c8d260c..9fa0bfb 100644
--- a/hw/vga-pci.c
+++ b/hw/vga-pci.c
@@ -79,6 +79,10 @@ static int pci_vga_initfn(PCIDevice *dev)
      VGACommonState *s = &d->vga;
      uint8_t *pci_conf = d->dev.config;
 
+     if (is_ioport_assigned(0x3c0)) {
+         return -1;
+     }
+
      // vga + console init
      vga_common_init(s, VGA_RAM_SIZE);
      vga_init(s);
diff --git a/hw/vmware_vga.c b/hw/vmware_vga.c
index 4e7a75d..7a8cec0 100644
--- a/hw/vmware_vga.c
+++ b/hw/vmware_vga.c
@@ -1232,6 +1232,10 @@ static int pci_vmsvga_initfn(PCIDevice *dev)
     struct pci_vmsvga_state_s *s =
         DO_UPCAST(struct pci_vmsvga_state_s, card, dev);
 
+    if (is_ioport_assigned(0x3c0)) {
+        return -1;
+    }
+
     pci_config_set_vendor_id(s->card.config, PCI_VENDOR_ID_VMWARE);
     pci_config_set_device_id(s->card.config, SVGA_PCI_DEVICE_ID);
     s->card.config[PCI_COMMAND]	= PCI_COMMAND_IO |
-- 
1.6.6.1

Re: [Qemu-devel] [PATCH 1/2] ioport: add function to check whenever a port is assigned or not

[Paul Brook]

> +int is_ioport_assigned(pio_addr_t addr)

Shouldn't we move this into register_ioport_{read,write}, and have that fail 
if the port has already been assigned?

Paul

Re: [Qemu-devel] [PATCH 1/2] ioport: add function to check whenever a port is assigned or not

[Gerd Hoffmann]

On 05/24/10 14:32, Paul Brook wrote:
>> +int is_ioport_assigned(pio_addr_t addr)
>
> Shouldn't we move this into register_ioport_{read,write}, and have that fail
> if the port has already been assigned?

It already checks and fails with hw_error().  Problem with that is that 
this kills qemu in case you try to pci hot-plug a vga card.  So I've 
added a way to check before-hand, so we can fail gracefully in the few 
places where we need it (see second patch of the series).

cheers,
   Gerd

Re: [Qemu-devel] [PATCH 1/2] ioport: add function to check whenever a port is assigned or not

[Markus Armbruster]

Gerd Hoffmann <kraxel@redhat.com> writes:

> On 05/24/10 14:32, Paul Brook wrote:
>>> +int is_ioport_assigned(pio_addr_t addr)
>>
>> Shouldn't we move this into register_ioport_{read,write}, and have that fail
>> if the port has already been assigned?
>
> It already checks and fails with hw_error().  Problem with that is
> that this kills qemu in case you try to pci hot-plug a vga card.  So
> I've added a way to check before-hand, so we can fail gracefully in
> the few places where we need it (see second patch of the series).

If we needed in more than a few places, then a solution like the one we
adopted for qdev_init() could make sense: have register_ioport_FOO()
return an error, convert all users that don't want to check it to
register_ioport_FOO_nofail(), which hw_error()s out.

[Qemu-devel] [PATCH 1/2] ioport: add function to check whenever a port is assigned or not

[Gerd Hoffmann]

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
---
 ioport.c |    5 +++++
 ioport.h |    1 +
 2 files changed, 6 insertions(+), 0 deletions(-)

diff --git a/ioport.c b/ioport.c
index 53dd87a..b718047 100644
--- a/ioport.c
+++ b/ioport.c
@@ -190,6 +190,11 @@ void isa_unassign_ioport(pio_addr_t start, int length)
     }
 }
 
+int is_ioport_assigned(pio_addr_t addr)
+{
+    return ioport_opaque[addr] != NULL;
+}
+
 /***********************************************************/
 
 void cpu_outb(pio_addr_t addr, uint8_t val)
diff --git a/ioport.h b/ioport.h
index 3d3c8a3..46fbfa8 100644
--- a/ioport.h
+++ b/ioport.h
@@ -41,6 +41,7 @@ int register_ioport_read(pio_addr_t start, int length, int size,
 int register_ioport_write(pio_addr_t start, int length, int size,
                           IOPortWriteFunc *func, void *opaque);
 void isa_unassign_ioport(pio_addr_t start, int length);
+int is_ioport_assigned(pio_addr_t addr);
 
 
 void cpu_outb(pio_addr_t addr, uint8_t val);
-- 
1.6.6.1

[Qemu-devel] [PATCH 2/2] all vga: fail graicefully when vga ports are taken

[Gerd Hoffmann]

Try to pci hotplug a vga card, watch qemu die with hw_error().
This patch fixes it.

Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
---
 hw/cirrus_vga.c |    3 +++
 hw/vga-pci.c    |    3 +++
 hw/vmware_vga.c |    3 +++
 3 files changed, 9 insertions(+), 0 deletions(-)

diff --git a/hw/cirrus_vga.c b/hw/cirrus_vga.c
index 9f61a01..2f50e86 100644
--- a/hw/cirrus_vga.c
+++ b/hw/cirrus_vga.c
@@ -3189,6 +3189,9 @@ static int pci_cirrus_vga_initfn(PCIDevice *dev)
      uint8_t *pci_conf = d->dev.config;
      int device_id = CIRRUS_ID_CLGD5446;
 
+     if (is_ioport_assigned(0x3c0))
+         return -1;
+
      /* setup VGA */
      vga_common_init(&s->vga, VGA_RAM_SIZE);
      cirrus_init_common(s, device_id, 1);
diff --git a/hw/vga-pci.c b/hw/vga-pci.c
index c8d260c..2ef3278 100644
--- a/hw/vga-pci.c
+++ b/hw/vga-pci.c
@@ -79,6 +79,9 @@ static int pci_vga_initfn(PCIDevice *dev)
      VGACommonState *s = &d->vga;
      uint8_t *pci_conf = d->dev.config;
 
+     if (is_ioport_assigned(0x3c0))
+         return -1;
+
      // vga + console init
      vga_common_init(s, VGA_RAM_SIZE);
      vga_init(s);
diff --git a/hw/vmware_vga.c b/hw/vmware_vga.c
index 4e7a75d..cd0eb57 100644
--- a/hw/vmware_vga.c
+++ b/hw/vmware_vga.c
@@ -1232,6 +1232,9 @@ static int pci_vmsvga_initfn(PCIDevice *dev)
     struct pci_vmsvga_state_s *s =
         DO_UPCAST(struct pci_vmsvga_state_s, card, dev);
 
+    if (is_ioport_assigned(0x3c0))
+        return -1;
+
     pci_config_set_vendor_id(s->card.config, PCI_VENDOR_ID_VMWARE);
     pci_config_set_device_id(s->card.config, SVGA_PCI_DEVICE_ID);
     s->card.config[PCI_COMMAND]	= PCI_COMMAND_IO |
-- 
1.6.6.1

Re: [Qemu-devel] [PATCH 2/2] all vga: fail graicefully when vga ports are taken

[Blue Swirl]

On 5/11/10, Gerd Hoffmann <kraxel@redhat.com> wrote:
> Try to pci hotplug a vga card, watch qemu die with hw_error().
>  This patch fixes it.
>
>  Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
>  ---
>   hw/cirrus_vga.c |    3 +++
>   hw/vga-pci.c    |    3 +++
>   hw/vmware_vga.c |    3 +++
>   3 files changed, 9 insertions(+), 0 deletions(-)
>
>  diff --git a/hw/cirrus_vga.c b/hw/cirrus_vga.c
>  index 9f61a01..2f50e86 100644
>  --- a/hw/cirrus_vga.c
>  +++ b/hw/cirrus_vga.c
>  @@ -3189,6 +3189,9 @@ static int pci_cirrus_vga_initfn(PCIDevice *dev)
>       uint8_t *pci_conf = d->dev.config;
>       int device_id = CIRRUS_ID_CLGD5446;
>
>  +     if (is_ioport_assigned(0x3c0))
>  +         return -1;
>  +

-ECODING_STYLE.

>       /* setup VGA */
>       vga_common_init(&s->vga, VGA_RAM_SIZE);
>       cirrus_init_common(s, device_id, 1);
>  diff --git a/hw/vga-pci.c b/hw/vga-pci.c
>  index c8d260c..2ef3278 100644
>  --- a/hw/vga-pci.c
>  +++ b/hw/vga-pci.c
>  @@ -79,6 +79,9 @@ static int pci_vga_initfn(PCIDevice *dev)
>       VGACommonState *s = &d->vga;
>       uint8_t *pci_conf = d->dev.config;
>
>  +     if (is_ioport_assigned(0x3c0))
>  +         return -1;
>  +
>       // vga + console init
>       vga_common_init(s, VGA_RAM_SIZE);
>       vga_init(s);
>  diff --git a/hw/vmware_vga.c b/hw/vmware_vga.c
>  index 4e7a75d..cd0eb57 100644
>  --- a/hw/vmware_vga.c
>  +++ b/hw/vmware_vga.c
>  @@ -1232,6 +1232,9 @@ static int pci_vmsvga_initfn(PCIDevice *dev)
>      struct pci_vmsvga_state_s *s =
>          DO_UPCAST(struct pci_vmsvga_state_s, card, dev);
>
>  +    if (is_ioport_assigned(0x3c0))
>  +        return -1;
>  +
>      pci_config_set_vendor_id(s->card.config, PCI_VENDOR_ID_VMWARE);
>      pci_config_set_device_id(s->card.config, SVGA_PCI_DEVICE_ID);
>      s->card.config[PCI_COMMAND]        = PCI_COMMAND_IO |
>
> --
>  1.6.6.1
>
>
>

Re: [Qemu-devel] [PATCH 2/2] all vga: fail graicefully when vga ports are taken

[Markus Armbruster]

Gerd Hoffmann <kraxel@redhat.com> writes:

> Try to pci hotplug a vga card, watch qemu die with hw_error().
> This patch fixes it.

Looks good.

Are there any other hot-pluggable devices that acquire single-use
resources such as fixed I/O ports?

Re: [Qemu-devel] [PATCH 2/2] all vga: fail graicefully when vga ports are taken

[Gerd Hoffmann]

On 05/18/10 14:33, Markus Armbruster wrote:
> Gerd Hoffmann<kraxel@redhat.com>  writes:
>
>> Try to pci hotplug a vga card, watch qemu die with hw_error().
>> This patch fixes it.
>
> Looks good.
>
> Are there any other hot-pluggable devices that acquire single-use
> resources such as fixed I/O ports?

Any PCI device doing ISA compatibility stuff I'd guess, which makes IDE 
a candidate.  Checking ...

Yes (hw/ide/piix.c):

     ide_init_ioport(&d->bus[0], 0x1f0, 0x3f6);
     ide_init_ioport(&d->bus[1], 0x170, 0x376);

Those are tagged with "no-user" though due to being hard-coded in pc.c, 
so any attempt to hot-plug one of those should fail way before it 
attempts to grab the I/O ports.

cheers,
   Gerd

Re: [Qemu-devel] [PATCH 2/2] all vga: fail graicefully when vga ports are taken

[Markus Armbruster]

Gerd Hoffmann <kraxel@redhat.com> writes:

> On 05/18/10 14:33, Markus Armbruster wrote:
>> Gerd Hoffmann<kraxel@redhat.com>  writes:
>>
>>> Try to pci hotplug a vga card, watch qemu die with hw_error().
>>> This patch fixes it.
>>
>> Looks good.
>>
>> Are there any other hot-pluggable devices that acquire single-use
>> resources such as fixed I/O ports?
>
> Any PCI device doing ISA compatibility stuff I'd guess, which makes
> IDE a candidate.  Checking ...
>
> Yes (hw/ide/piix.c):
>
>     ide_init_ioport(&d->bus[0], 0x1f0, 0x3f6);
>     ide_init_ioport(&d->bus[1], 0x170, 0x376);
>
> Those are tagged with "no-user" though due to being hard-coded in
> pc.c, so any attempt to hot-plug one of those should fail way before
> it attempts to grab the I/O ports.

So a more general mechanism than you ad hoc is_ioport_assigned() check
doesn't seem to be justified.