[9fans] The late upamalloc

[9fans] The late upamalloc

[Brian L. Stuart]

Okay, I've been taking a crack at the broadcom driver
that's been lying around.  It's pretty old and used
the upamalloc call.  But upamalloc is now gone and
I don't seem to be able to find the right way to
replace it.  After looking at other drivers and at
the old upamalloc code on sourcesdump, I've tried
several way to replace it, but most end up in a panic.
The original code (edited to the relevant lines)
looked like:

port = upamalloc(bar & ~0x0F, p->mem[0].size, 0);
ctlr->port = port;
ctlr->nic = KADDR(ctlr->port);

My most recent attempt replaced the upamalloc call
with:

upareserve(bar & ~0x0F, p->mem[0].size);
pdbmap(MACH(0)->pdb, bar & ~0x0F, 0, p->mem[0].size);
port = bar & ~0x0F;

There is currently a upaalloc call, and the comment
to it seems to imply that vmap should be used either
in conjunction with upaalloc or by it self.  I've
tried a few permutations of that with no luck.

If someone knows off the top of their head the right
way to replace an old upamalloc call, it would save
me some time, and I could get on to seeing if I can
manage to get the controller to work.

Thanks in advance,
BLS

Re: [9fans] The late upamalloc

[erik quanstrom]

uh, cut-n-paste error... that should be

	ctlr->nic = vmap(p->mem[0].bar & ~0xf, p->mem[0].size);
	ctlr->port = PCIWADDR(ctlr->nic);

- erik

Re: [9fans] The late upamalloc

[Uriel]

If you are working on the driver in my contrib dir, there are probably
newer versions of it somewhere, but don't ask me where you could find
them or who might have them, I don't even remember where I got that
one from.

As far as I know some of the broadcom work was done together with the
amd64 port, so good luck getting hold of that, hahaha.

Pace

uriel

On Sat, Sep 27, 2008 at 5:43 PM, Brian L. Stuart <blstuart@bellsouth.net> wrote:
> Okay, I've been taking a crack at the broadcom driver
> that's been lying around.  It's pretty old and used
> the upamalloc call.  But upamalloc is now gone and
> I don't seem to be able to find the right way to
> replace it.  After looking at other drivers and at
> the old upamalloc code on sourcesdump, I've tried
> several way to replace it, but most end up in a panic.
> The original code (edited to the relevant lines)
> looked like:
>
> port = upamalloc(bar & ~0x0F, p->mem[0].size, 0);
> ctlr->port = port;
> ctlr->nic = KADDR(ctlr->port);
>
> My most recent attempt replaced the upamalloc call
> with:
>
> upareserve(bar & ~0x0F, p->mem[0].size);
> pdbmap(MACH(0)->pdb, bar & ~0x0F, 0, p->mem[0].size);
> port = bar & ~0x0F;
>
> There is currently a upaalloc call, and the comment
> to it seems to imply that vmap should be used either
> in conjunction with upaalloc or by it self.  I've
> tried a few permutations of that with no luck.
>
> If someone knows off the top of their head the right
> way to replace an old upamalloc call, it would save
> me some time, and I could get on to seeing if I can
> manage to get the controller to work.
>
> Thanks in advance,
> BLS
>
>

Re: [9fans] The late upamalloc

[Brian L. Stuart]

> If you are working on the driver in my contrib dir, there are probably
> newer versions of it somewhere, but don't ask me where you could find
> them or who might have them, I don't even remember where I got that
> one from.

I think that's the one I started with.

> As far as I know some of the broadcom work was done together with the
> amd64 port, so good luck getting hold of that, hahaha.

Jim sent me one that's a little farther along, but not a lot
farther.  It did get me to the point where I can initialize
the controller.  But I've still got a ways to go on getting
it to actually communicate.

Thanks,
BLS

Re: [9fans] The late upamalloc

[erik quanstrom]

> Okay, I've been taking a crack at the broadcom driver
> that's been lying around.  It's pretty old and used
> the upamalloc call.  But upamalloc is now gone and
> I don't seem to be able to find the right way to
> replace it.  After looking at other drivers and at
> the old upamalloc code on sourcesdump, I've tried
> several way to replace it, but most end up in a panic.
> The original code (edited to the relevant lines)
> looked like:
>
> port = upamalloc(bar & ~0x0F, p->mem[0].size, 0);
> ctlr->port = port;
> ctlr->nic = KADDR(ctlr->port);
>
> My most recent attempt replaced the upamalloc call
> with:
>
> upareserve(bar & ~0x0F, p->mem[0].size);
> pdbmap(MACH(0)->pdb, bar & ~0x0F, 0, p->mem[0].size);
> port = bar & ~0x0F;
>
> There is currently a upaalloc call, and the comment
> to it seems to imply that vmap should be used either
> in conjunction with upaalloc or by it self.  I've
> tried a few permutations of that with no luck.
>
> If someone knows off the top of their head the right
> way to replace an old upamalloc call, it would save
> me some time, and I could get on to seeing if I can
> manage to get the controller to work.
>

upamalloc should be replaced with vmap.  but be careful.
the address returned is a virtual kernel address.  so anyone
assuming that the address returned is going to be a physical
address is going to loose.  i think you wish

	ctlr->nic = vmap(p->mem[Abar].bar & ~0xf, p->mem[0].size);
	ctlr->port = PCIWADDR(ctlr->nic);

- erik