2005-04-13 21:31:52

by Bjorn Helgaas

[permalink] [raw]
Subject: [PATCH] PC300 pci_enable_device fix

Call pci_enable_device() before looking at IRQ and resources.
The driver requires this fix or the "pci=routeirq" workaround
on 2.6.10 and later kernels.

Reported and tested by Artur Lipowski.

Signed-off-by: Bjorn Helgaas <[email protected]>

===== drivers/net/wan/pc300_drv.c 1.24 vs edited =====
--- 1.24/drivers/net/wan/pc300_drv.c 2004-12-29 12:25:16 -07:00
+++ edited/drivers/net/wan/pc300_drv.c 2005-04-13 13:35:21 -06:00
@@ -3439,6 +3439,9 @@
#endif
}

+ if ((err = pci_enable_device(pdev)) != 0)
+ return err;
+
card = (pc300_t *) kmalloc(sizeof(pc300_t), GFP_KERNEL);
if (card == NULL) {
printk("PC300 found at RAM 0x%08lx, "
@@ -3526,9 +3529,6 @@
err = -ENODEV;
goto err_release_ram;
}
-
- if ((err = pci_enable_device(pdev)) != 0)
- goto err_release_sca;

card->hw.plxbase = ioremap(card->hw.plxphys, card->hw.plxsize);
card->hw.rambase = ioremap(card->hw.ramphys, card->hw.alloc_ramsize);



2005-04-13 22:02:58

by Ashok Raj

[permalink] [raw]
Subject: Re: [PATCH] PC300 pci_enable_device fix

On Wed, Apr 13, 2005 at 02:31:43PM -0700, Bjorn Helgaas wrote:
>
> Call pci_enable_device() before looking at IRQ and resources.
> The driver requires this fix or the "pci=routeirq" workaround
> on 2.6.10 and later kernels.


the failure cases dont seem to worry about pci_disable_device()?

in err_release_ram: etc?

>
> Reported and tested by Artur Lipowski.
>
> Signed-off-by: Bjorn Helgaas <[email protected]>
>
> ===== drivers/net/wan/pc300_drv.c 1.24 vs edited =====
> --- 1.24/drivers/net/wan/pc300_drv.c 2004-12-29 12:25:16 -07:00
> +++ edited/drivers/net/wan/pc300_drv.c 2005-04-13 13:35:21 -06:00
> @@ -3439,6 +3439,9 @@
> #endif
> }
>
> + if ((err = pci_enable_device(pdev)) != 0)
> + return err;
> +
> card = (pc300_t *) kmalloc(sizeof(pc300_t), GFP_KERNEL);
> if (card == NULL) {
> printk("PC300 found at RAM 0x%08lx, "
> @@ -3526,9 +3529,6 @@
> err = -ENODEV;
> goto err_release_ram;
> }
> -
> - if ((err = pci_enable_device(pdev)) != 0)
> - goto err_release_sca;
>
> card->hw.plxbase = ioremap(card->hw.plxphys,
> card->hw.plxsize);
> card->hw.rambase = ioremap(card->hw.ramphys,
> card->hw.alloc_ramsize);
>
> -
> To unsubscribe from this list: send the line "unsubscribe
> linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at [1]http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at [2]http://www.tux.org/lkml/
>
> References
>
> 1. http://vger.kernel.org/majordomo-info.html
> 2. http://www.tux.org/lkml/

--
Cheers,
Ashok Raj
- Open Source Technology Center

2005-04-14 00:43:38

by Marcelo Tosatti

[permalink] [raw]
Subject: Re: [PATCH] PC300 pci_enable_device fix

On Wed, Apr 13, 2005 at 03:02:43PM -0700, Ashok Raj wrote:
> On Wed, Apr 13, 2005 at 02:31:43PM -0700, Bjorn Helgaas wrote:
> >
> > Call pci_enable_device() before looking at IRQ and resources.
> > The driver requires this fix or the "pci=routeirq" workaround
> > on 2.6.10 and later kernels.
>
>
> the failure cases dont seem to worry about pci_disable_device()?
>
> in err_release_ram: etc?

Yep the failure paths were wrong before, but Bjorn's patch moves
pci_enable_device() way up to the beginning of the function.

The failure path's err_release_ram etc. wont touch the resources
without pci_enable_pci(), with the fix.

> > Reported and tested by Artur Lipowski.
> >
> > Signed-off-by: Bjorn Helgaas <[email protected]>
> >
> > ===== drivers/net/wan/pc300_drv.c 1.24 vs edited =====
> > --- 1.24/drivers/net/wan/pc300_drv.c 2004-12-29 12:25:16 -07:00
> > +++ edited/drivers/net/wan/pc300_drv.c 2005-04-13 13:35:21 -06:00
> > @@ -3439,6 +3439,9 @@
> > #endif
> > }
> >
> > + if ((err = pci_enable_device(pdev)) != 0)
> > + return err;
> > +
> > card = (pc300_t *) kmalloc(sizeof(pc300_t), GFP_KERNEL);
> > if (card == NULL) {
> > printk("PC300 found at RAM 0x%08lx, "
> > @@ -3526,9 +3529,6 @@
> > err = -ENODEV;
> > goto err_release_ram;
> > }
> > -
> > - if ((err = pci_enable_device(pdev)) != 0)
> > - goto err_release_sca;
> >
> > card->hw.plxbase = ioremap(card->hw.plxphys,
> > card->hw.plxsize);
> > card->hw.rambase = ioremap(card->hw.ramphys,
> > card->hw.alloc_ramsize);

2005-04-14 18:18:09

by Bjorn Helgaas

[permalink] [raw]
Subject: Re: [PATCH] PC300 pci_enable_device fix

On Wed, 2005-04-13 at 15:02 -0700, Ashok Raj wrote:
> On Wed, Apr 13, 2005 at 02:31:43PM -0700, Bjorn Helgaas wrote:
> >
> > Call pci_enable_device() before looking at IRQ and resources.
> > The driver requires this fix or the "pci=routeirq" workaround
> > on 2.6.10 and later kernels.

> the failure cases dont seem to worry about pci_disable_device()?
> in err_release_ram: etc?

True. The failure paths were broken before, but here's a
revised patch that adds pci_disable_device() to them as
well. Artur has tested this patch, too.


Call pci_enable_device() before looking at IRQ and resources.
The driver requires this fix or the "pci=routeirq" workaround
on 2.6.10 and later kernels.

Reported and tested by Artur Lipowski.

Signed-off-by: Bjorn Helgaas <[email protected]>

===== drivers/net/wan/pc300_drv.c 1.24 vs edited =====
--- 1.24/drivers/net/wan/pc300_drv.c 2004-12-29 12:25:16 -07:00
+++ edited/drivers/net/wan/pc300_drv.c 2005-04-13 16:14:19 -06:00
@@ -3427,7 +3427,7 @@
{
static int first_time = 1;
ucchar cpc_rev_id;
- int err = 0, eeprom_outdated = 0;
+ int err, eeprom_outdated = 0;
ucshort device_id;
pc300_t *card;

@@ -3439,15 +3439,21 @@
#endif
}

+ if ((err = pci_enable_device(pdev)) < 0)
+ return err;
+
card = (pc300_t *) kmalloc(sizeof(pc300_t), GFP_KERNEL);
if (card == NULL) {
printk("PC300 found at RAM 0x%08lx, "
"but could not allocate card structure.\n",
pci_resource_start(pdev, 3));
- return -ENOMEM;
+ err = -ENOMEM;
+ goto err_disable_dev;
}
memset(card, 0, sizeof(pc300_t));

+ err = -ENODEV;
+
/* read PCI configuration area */
device_id = ent->device;
card->hw.irq = pdev->irq;
@@ -3507,7 +3513,6 @@
printk("PC300 found at RAM 0x%08x, "
"but could not allocate PLX mem region.\n",
card->hw.ramphys);
- err = -ENODEV;
goto err_release_io;
}
if (!request_mem_region(card->hw.ramphys, card->hw.alloc_ramsize,
@@ -3515,7 +3520,6 @@
printk("PC300 found at RAM 0x%08x, "
"but could not allocate RAM mem region.\n",
card->hw.ramphys);
- err = -ENODEV;
goto err_release_plx;
}
if (!request_mem_region(card->hw.scaphys, card->hw.scasize,
@@ -3523,13 +3527,9 @@
printk("PC300 found at RAM 0x%08x, "
"but could not allocate SCA mem region.\n",
card->hw.ramphys);
- err = -ENODEV;
goto err_release_ram;
}

- if ((err = pci_enable_device(pdev)) != 0)
- goto err_release_sca;
-
card->hw.plxbase = ioremap(card->hw.plxphys, card->hw.plxsize);
card->hw.rambase = ioremap(card->hw.ramphys, card->hw.alloc_ramsize);
card->hw.scabase = ioremap(card->hw.scaphys, card->hw.scasize);
@@ -3619,7 +3619,6 @@
iounmap(card->hw.falcbase);
release_mem_region(card->hw.falcphys, card->hw.falcsize);
}
-err_release_sca:
release_mem_region(card->hw.scaphys, card->hw.scasize);
err_release_ram:
release_mem_region(card->hw.ramphys, card->hw.alloc_ramsize);
@@ -3628,7 +3627,9 @@
err_release_io:
release_region(card->hw.iophys, card->hw.iosize);
kfree(card);
- return -ENODEV;
+err_disable_dev:
+ pci_disable_device(pdev);
+ return err;
}

static void __devexit cpc_remove_one(struct pci_dev *pdev)


2005-04-14 18:32:20

by Marcelo Tosatti

[permalink] [raw]
Subject: Re: [PATCH] PC300 pci_enable_device fix


Germano, can you give this patch a try before merging, please.

On Wed, Apr 13, 2005 at 04:45:31PM -0300, Marcelo Tosatti wrote:
> On Wed, Apr 13, 2005 at 03:02:43PM -0700, Ashok Raj wrote:
> > On Wed, Apr 13, 2005 at 02:31:43PM -0700, Bjorn Helgaas wrote:
> > >
> > > Call pci_enable_device() before looking at IRQ and resources.
> > > The driver requires this fix or the "pci=routeirq" workaround
> > > on 2.6.10 and later kernels.
> >
> >
> > the failure cases dont seem to worry about pci_disable_device()?
> >
> > in err_release_ram: etc?
>
> Yep the failure paths were wrong before, but Bjorn's patch moves
> pci_enable_device() way up to the beginning of the function.
>
> The failure path's err_release_ram etc. wont touch the resources
> without pci_enable_pci(), with the fix.

Hi,

I see what you mean now, Ashok. Yes, you are right, the error path lacks
pci_disable_device().

Reported by Artur Lipowski.

Signed-off-by: Bjorn Helgaas <[email protected]>
Signed-off-by: Marcelo Tosatti <[email protected]>

--- a/drivers/net/wan/pc300_drv.c.orig 2005-04-14 13:56:04.000000000 -0300
+++ b/drivers/net/wan/pc300_drv.c 2005-04-14 13:58:56.000000000 -0300
@@ -3439,11 +3439,15 @@
#endif
}

+ if ((err = pci_enable_device(pdev)) != 0)
+ return err;
+
card = (pc300_t *) kmalloc(sizeof(pc300_t), GFP_KERNEL);
if (card == NULL) {
printk("PC300 found at RAM 0x%08lx, "
"but could not allocate card structure.\n",
pci_resource_start(pdev, 3));
+ pci_disable_device(pdev);
return -ENOMEM;
}
memset(card, 0, sizeof(pc300_t));
@@ -3527,8 +3531,6 @@
goto err_release_ram;
}

- if ((err = pci_enable_device(pdev)) != 0)
- goto err_release_sca;

card->hw.plxbase = ioremap(card->hw.plxphys, card->hw.plxsize);
card->hw.rambase = ioremap(card->hw.ramphys, card->hw.alloc_ramsize);
@@ -3628,6 +3630,7 @@
err_release_io:
release_region(card->hw.iophys, card->hw.iosize);
kfree(card);
+ pci_disable_device(pdev);
return -ENODEV;
}