Olaf Hering <[EMAIL PROTECTED]> writes:
> - printk(KERN_ERR "radeonfb (%s): Invalid ROM signature %x should
> be"
> + printk(KERN_DEBUG "radeonfb (%s): Invalid ROM signature %x
> should be"
> "0xaa55\n", pci_name(rinfo->pdev), BIOS_IN16(0));
While yo
On Tue, Feb 15, Jesse Barnes wrote:
> Both the r128 and radeon drivers complain if they don't find an x86 option
> ROM
> on the device they're talking to. This would be fine, except that the
> message is incorrect--not all option ROMs are required to be x86 based. This
> small patch just re
On Fri, 18 Feb 2005 13:09:14 +0100, Gabriel Paubert <[EMAIL PROTECTED]> wrote:
> For example if it declares 128k, compare the two halves, reduce
> to 64k if equal. Lather, rinse, repeat.
>
> It's equivalent to reading the BAR declared size twice in
> the worst case, so it's not that bad performanc
On Thu, Feb 17, 2005 at 05:56:03PM -0500, Jon Smirl wrote:
> On Fri, 18 Feb 2005 09:45:50 +1100, Benjamin Herrenschmidt
> <[EMAIL PROTECTED]> wrote:
>
> > Can't the size be obtained like any other BAR ?
>
> yes, but cards that don't fully decode their ROM address space can
> waste memory in copy_
On Thu, 17 Feb 2005, Jon Smirl wrote:
> On Fri, 18 Feb 2005 09:47:15 +1100, Benjamin Herrenschmidt
> <[EMAIL PROTECTED]> wrote:
> > We could provide additional helpers, like pci_find_rom_partition(),
> > which takes the architecture code as an argument. It would check the
> > signature, and iterate
On Thu, 2005-02-17 at 17:59 -0500, Jon Smirl wrote:
> On Fri, 18 Feb 2005 09:47:15 +1100, Benjamin Herrenschmidt
> <[EMAIL PROTECTED]> wrote:
> > We could provide additional helpers, like pci_find_rom_partition(),
> > which takes the architecture code as an argument. It would check the
> > signatur
On Thursday, February 17, 2005 2:59 pm, Jon Smirl wrote:
> On Fri, 18 Feb 2005 09:47:15 +1100, Benjamin Herrenschmidt
>
> <[EMAIL PROTECTED]> wrote:
> > We could provide additional helpers, like pci_find_rom_partition(),
> > which takes the architecture code as an argument. It would check the
> > s
On Fri, 18 Feb 2005 09:47:15 +1100, Benjamin Herrenschmidt
<[EMAIL PROTECTED]> wrote:
> We could provide additional helpers, like pci_find_rom_partition(),
> which takes the architecture code as an argument. It would check the
> signature, and iterate all "partitions" til it finds the proper
> arch
On Fri, 18 Feb 2005 09:45:50 +1100, Benjamin Herrenschmidt
<[EMAIL PROTECTED]> wrote:
> Can't the size be obtained like any other BAR ?
yes, but cards that don't fully decode their ROM address space can
waste memory in copy_rom. For example I have a card around here that
reports a BAR address spa
On Thu, 2005-02-17 at 12:56 -0500, Jon Smirl wrote:
> On Thu, 17 Feb 2005 09:45:30 -0800, Jesse Barnes <[EMAIL PROTECTED]> wrote:
> > Ok, how does this one look to you guys? The r128 driver would need similar
> > fixes.
>
> Do any of the radeon ROMs store multiple images in different formats?
> S
On Thu, 2005-02-17 at 09:45 -0800, Jesse Barnes wrote:
> On Thursday, February 17, 2005 9:32 am, Jon Smirl wrote:
> > On Thu, 17 Feb 2005 09:29:53 -0800, Jesse Barnes <[EMAIL PROTECTED]> wrote:
> > > On Thursday, February 17, 2005 8:33 am, Jon Smirl wrote:
> > > > > No, pci_map_rom shouldn't test t
On Thu, 2005-02-17 at 11:33 -0500, Jon Smirl wrote:
> On Thu, 17 Feb 2005 11:48:14 +1100, Benjamin Herrenschmidt
> <[EMAIL PROTECTED]> wrote:
> > On Wed, 2005-02-16 at 15:54 -0800, Jesse Barnes wrote:
> > > On Tuesday, February 15, 2005 5:03 pm, Benjamin Herrenschmidt wrote:
> > > > What about prin
On Thu, 17 Feb 2005 09:45:30 -0800, Jesse Barnes <[EMAIL PROTECTED]> wrote:
> Ok, how does this one look to you guys? The r128 driver would need similar
> fixes.
Do any of the radeon ROMs store multiple images in different formats?
Should the radeon driver loop throught the ROM images looking for
On Thursday, February 17, 2005 9:32 am, Jon Smirl wrote:
> On Thu, 17 Feb 2005 09:29:53 -0800, Jesse Barnes <[EMAIL PROTECTED]> wrote:
> > On Thursday, February 17, 2005 8:33 am, Jon Smirl wrote:
> > > > No, pci_map_rom shouldn't test the signature IMHO. While PCI ROMs
> > > > should have the signa
On Thursday, February 17, 2005 9:32 am, Jon Smirl wrote:
> > Shouldn't it return NULL if the signature is invalid?
>
> But then you couldn't get to your non-standard ROMs
Ok, I'll fix up the callers then.
Jesse
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body
On Thu, 17 Feb 2005 09:29:53 -0800, Jesse Barnes <[EMAIL PROTECTED]> wrote:
> On Thursday, February 17, 2005 8:33 am, Jon Smirl wrote:
> > > No, pci_map_rom shouldn't test the signature IMHO. While PCI ROMs should
> > > have the signature to be recognized as containing valid firmware images
> > > o
On Thursday, February 17, 2005 8:33 am, Jon Smirl wrote:
> > No, pci_map_rom shouldn't test the signature IMHO. While PCI ROMs should
> > have the signature to be recognized as containing valid firmware images
> > on x86 BIOSes an OF, it's just a convention on these platforms, and I
> > would rathe
On Thu, 17 Feb 2005 11:48:14 +1100, Benjamin Herrenschmidt
<[EMAIL PROTECTED]> wrote:
> On Wed, 2005-02-16 at 15:54 -0800, Jesse Barnes wrote:
> > On Tuesday, February 15, 2005 5:03 pm, Benjamin Herrenschmidt wrote:
> > > What about printing "No PCI ROM detected" ? I like having that info when
> >
On Wed, 2005-02-16 at 15:54 -0800, Jesse Barnes wrote:
> On Tuesday, February 15, 2005 5:03 pm, Benjamin Herrenschmidt wrote:
> > What about printing "No PCI ROM detected" ? I like having that info when
> > getting user reports, but I agree that a less worrying message would
> > be good.
>
> Ok, h
On Tuesday, February 15, 2005 5:03 pm, Benjamin Herrenschmidt wrote:
> What about printing "No PCI ROM detected" ? I like having that info when
> getting user reports, but I agree that a less worrying message would
> be good.
Ok, how about this then? It changes the printks in both drivers to KERN
On Tue, 2005-02-15 at 20:08 -0500, Jon Smirl wrote:
> There is a new io resource flag as part of the pci rom code,
> IORESOURCE_ROM_SHADOW, that is used on x86. If IORESOURCE_ROM_SHADOW
> is set, you should ignore the physical ROM and use the copy at C000:0.
> Can we build an equivalent flag for PP
On Tuesday, February 15, 2005 5:08 pm, Jon Smirl wrote:
> There is a new io resource flag as part of the pci rom code,
> IORESOURCE_ROM_SHADOW, that is used on x86. If IORESOURCE_ROM_SHADOW
> is set, you should ignore the physical ROM and use the copy at C000:0.
> Can we build an equivalent flag fo
There is a new io resource flag as part of the pci rom code,
IORESOURCE_ROM_SHADOW, that is used on x86. If IORESOURCE_ROM_SHADOW
is set, you should ignore the physical ROM and use the copy at C000:0.
Can we build an equivalent flag for PPC? On x86 arch specific code
determines the boot video devic
On Tue, 2005-02-15 at 15:57 -0800, Jesse Barnes wrote:
> Both the r128 and radeon drivers complain if they don't find an x86 option
> ROM
> on the device they're talking to. This would be fine, except that the
> message is incorrect--not all option ROMs are required to be x86 based. This
> sm
> I thought the signature described what type of ROM was there? E.g. 0xaa55
> means x86 ROM, x0303 means OF ROM, etc.?
>
> At any rate, not having a ROM at all (which my case may be) isn't an error
> either, so I think removing the printk is appropriate regardless.
Oh, and if this is the Powe
On Tue, 2005-02-15 at 16:45 -0800, Jesse Barnes wrote:
> On Tuesday, February 15, 2005 4:36 pm, Jon Smirl wrote:
> > You're removing the check for 55AA at the start of the ROM.
>
> No, the check is still there, I just removed the printk if 0xaa55 isn't found
> (my box returns 0x303 instead).
>
>
On Tuesday, February 15, 2005 4:36 pm, Jon Smirl wrote:
> You're removing the check for 55AA at the start of the ROM.
No, the check is still there, I just removed the printk if 0xaa55 isn't found
(my box returns 0x303 instead).
> I though
> the PCI standard was that all ROMs had to start with t
You're removing the check for 55AA at the start of the ROM. I though
the PCI standard was that all ROMs had to start with the no matter
what object code they contain. Then if you look for PCIR there is a
field in the stucture that says what language the ROM is in. Maybe the
problem is in the BIOS_I
Both the r128 and radeon drivers complain if they don't find an x86 option ROM
on the device they're talking to. This would be fine, except that the
message is incorrect--not all option ROMs are required to be x86 based. This
small patch just removes the messages altogether, causing the driver
29 matches
Mail list logo
