[Qemu-devel] Using ISA-PNP for resource management

[Qemu-devel] Using ISA-PNP for resource management

[Matthew Mastracci]

I was looking at how much work it would take to switch to using ISA-PNP
for resource management.  I believe that it might simplify some of the
QEMU internals, as well as make things like having multiple
network/other devices much easier.  It may make moving to PCI (or having
an option to boot with either ISAPNP or PCI on the fly) much easier.

For anyone interested, the ISAPNP spec page is at:

 http://osdev.neopages.net/docs/PNP-ISA-v1.0a.pdf?the_id=54

I believe the best way to implement ISA-PNP would be for each device to
register an ISA-PNP context, then add specific resource requirements to
the context as necessary.  The ISAPNP manager would then interact with
the operating system and call the isapnp_configure method whenever the
OS changes the setup.

Perhaps something like this.  I would imagine that the ISAPNP resource
manager would clear the IO port range back to the default IO port
manager before calling the *_isapnp_configure routine:

static void ne2000_isapnp_configure(void* opaque, ISAPNPContext *c)
{
    /* Get the first IO port as the base IO address */
    int base = isapnp_get_ioport(c, 0);

    register_ioport_write(base, 16, 1, ne2000_ioport_write, s);
    register_ioport_read(base, 16, 1, ne2000_ioport_read, s);

    register_ioport_write(base + 0x10, 1, 1, ne2000_asic_ioport_write,
s);
    register_ioport_read(base + 0x10, 1, 1, ne2000_asic_ioport_read, s);
    register_ioport_write(base + 0x10, 2, 2, ne2000_asic_ioport_write,
s);
    register_ioport_read(base + 0x10, 2, 2, ne2000_asic_ioport_read, s);

    register_ioport_write(base + 0x1f, 1, 1, ne2000_reset_ioport_write,
s);
    register_ioport_read(base + 0x1f, 1, 1, ne2000_reset_ioport_read,
s);

    s->irq = isapnp_get_irq(c, 0);

    /* Perhaps this would happen below... (?) */
    ne2000_reset();
}

void ne2000_init(int base, int irq, NetDriverState *nd)
{
    ISAPNPContext *c;
    /* KTC2000 is an ISAPNP vendor ID for an NE2000 compatible card */
    /* No serial number specified here, ISAPNP manager can autogenerate
it */
    c = isapnp_create_context("KTC", 2000, ne2000_isapnp_configure, s);

    /* Need an edge-triggered IRQ, default is irq */
    isapnp_register_irq(IRQ_MODE_EDGE, irq);

    /* Need a block of 32 IO ports, default is base */
    isapnp_register_ioport(32, base);
}
-- 
Matthew Mastracci <matt@aclaro.com>

Re: [Qemu-devel] Using ISA-PNP for resource management

[Fabrice Bellard]

Matthew Mastracci wrote:
> I was looking at how much work it would take to switch to using ISA-PNP
> for resource management.  I believe that it might simplify some of the
> QEMU internals, as well as make things like having multiple
> network/other devices much easier.  It may make moving to PCI (or having
> an option to boot with either ISAPNP or PCI on the fly) much easier.
> 
> For anyone interested, the ISAPNP spec page is at:
> 
>  http://osdev.neopages.net/docs/PNP-ISA-v1.0a.pdf?the_id=54
> 
> I believe the best way to implement ISA-PNP would be for each device to
> register an ISA-PNP context, then add specific resource requirements to
> the context as necessary.  The ISAPNP manager would then interact with
> the operating system and call the isapnp_configure method whenever the
> OS changes the setup.
> 
> Perhaps something like this.  I would imagine that the ISAPNP resource
> manager would clear the IO port range back to the default IO port
> manager before calling the *_isapnp_configure routine:

OK for ISAPNP, but it should just indicate what is the hardcoded 
configuration. I don't want to use it to set the exact ioports and irqs. 
Moreover PCI is about to be integrated in QEMU, so it may not be worth 
the effort.

Fabrice.

[Qemu-devel] Re: Using ISA-PNP for resource management

[Matthew Mastracci]

[-- Attachment #1: Type: text/plain, Size: 793 bytes --]

Fabrice Bellard wrote:

> OK for ISAPNP, but it should just indicate what is the hardcoded 
> configuration. I don't want to use it to set the exact ioports and 
> irqs. Moreover PCI is about to be integrated in QEMU, so it may not be 
> worth the effort.

Cool!  Will you be adding an option whether to include PCI or not in the 
host?  I believe that only devices actually on the ISA bus (not those on 
the PCI bus) will participate in the ISAPNP process.  If PCI is enabled, 
the ISAPNP enumerator should not respond (since no ISA devices are 
around to participate in the process). 

I suppose it might not be very useful to add ISAPNP if something as old 
as Windows '95 supports PCI.  PCI is far better for device enumeration 
anyways.  Might be better to wait in this case.  :)

Matt.


[-- Attachment #2: matt.vcf --]
[-- Type: text/x-vcard, Size: 286 bytes --]

begin:vcard
fn:Matthew Mastracci
n:Mastracci;Matthew
org:aclaro Softworks, inc.
adr:;;1900 a - 11 St. SE;Calgary;Alberta;T2H 3G2;Canada
email;internet:matt@aclaro.com
title:Software Developer
tel;work:(403) 299-6612
x-mozilla-html:FALSE
url:http://www.aclaro.com
version:2.1
end:vcard

[Qemu-devel] Re: Using ISA-PNP for resource management

[Matthew Mastracci]

[-- Attachment #1: Type: text/plain, Size: 2856 bytes --]

After playing around with this a bit, it might even be easier to 
register your IO port routines with the ISAPNP manager.  The ISAPNP 
manager could then add/remove them from the global IO port list as 
necessary:

    isapnp_register_ioport_write(c, 0, 16, 1, ne2000_ioport_write, s);
    isapnp_register_ioport_read(c, 0, 16, 1, ne2000_ioport_read, s);


Note that c is the ISAPNP context from below and 0 is the first bank of 
IO port registers.

Matt.

Matthew Mastracci wrote:

>I was looking at how much work it would take to switch to using ISA-PNP
>for resource management.  I believe that it might simplify some of the
>QEMU internals, as well as make things like having multiple
>network/other devices much easier.  It may make moving to PCI (or having
>an option to boot with either ISAPNP or PCI on the fly) much easier.
>
>For anyone interested, the ISAPNP spec page is at:
>
> http://osdev.neopages.net/docs/PNP-ISA-v1.0a.pdf?the_id=54
>
>I believe the best way to implement ISA-PNP would be for each device to
>register an ISA-PNP context, then add specific resource requirements to
>the context as necessary.  The ISAPNP manager would then interact with
>the operating system and call the isapnp_configure method whenever the
>OS changes the setup.
>
>Perhaps something like this.  I would imagine that the ISAPNP resource
>manager would clear the IO port range back to the default IO port
>manager before calling the *_isapnp_configure routine:
>
>static void ne2000_isapnp_configure(void* opaque, ISAPNPContext *c)
>{
>    /* Get the first IO port as the base IO address */
>    int base = isapnp_get_ioport(c, 0);
>
>    register_ioport_write(base, 16, 1, ne2000_ioport_write, s);
>    register_ioport_read(base, 16, 1, ne2000_ioport_read, s);
>
>    register_ioport_write(base + 0x10, 1, 1, ne2000_asic_ioport_write,
>s);
>    register_ioport_read(base + 0x10, 1, 1, ne2000_asic_ioport_read, s);
>    register_ioport_write(base + 0x10, 2, 2, ne2000_asic_ioport_write,
>s);
>    register_ioport_read(base + 0x10, 2, 2, ne2000_asic_ioport_read, s);
>
>    register_ioport_write(base + 0x1f, 1, 1, ne2000_reset_ioport_write,
>s);
>    register_ioport_read(base + 0x1f, 1, 1, ne2000_reset_ioport_read,
>s);
>
>    s->irq = isapnp_get_irq(c, 0);
>
>    /* Perhaps this would happen below... (?) */
>    ne2000_reset();
>}
>
>void ne2000_init(int base, int irq, NetDriverState *nd)
>{
>    ISAPNPContext *c;
>    /* KTC2000 is an ISAPNP vendor ID for an NE2000 compatible card */
>    /* No serial number specified here, ISAPNP manager can autogenerate
>it */
>    c = isapnp_create_context("KTC", 2000, ne2000_isapnp_configure, s);
>
>    /* Need an edge-triggered IRQ, default is irq */
>    isapnp_register_irq(IRQ_MODE_EDGE, irq);
>
>    /* Need a block of 32 IO ports, default is base */
>    isapnp_register_ioport(32, base);
>}
>  
>


[-- Attachment #2: matt.vcf --]
[-- Type: text/x-vcard, Size: 286 bytes --]

begin:vcard
fn:Matthew Mastracci
n:Mastracci;Matthew
org:aclaro Softworks, inc.
adr:;;1900 a - 11 St. SE;Calgary;Alberta;T2H 3G2;Canada
email;internet:matt@aclaro.com
title:Software Developer
tel;work:(403) 299-6612
x-mozilla-html:FALSE
url:http://www.aclaro.com
version:2.1
end:vcard