On Thu, Feb 21, 2008 at 03:47:33PM -0800, Greg Kroah-Hartman wrote:
> +static void __devinit quirk_amd_ide_mode(struct pci_dev *pdev)
> {
> - /* set sb600 sata to ahci mode */
> - if ((pdev->class >> 8) == PCI_CLASS_STORAGE_IDE) {
> - u8 tmp;
> + /* set sb600/sb700/sb800 sata to ahci mode */
> + u8 tmp;
>
> + pci_read_config_byte(pdev, PCI_CLASS_DEVICE, &tmp);
> + if (tmp == 0x01) {
> pci_read_config_byte(pdev, 0x40, &tmp);
This seems like a dis-improvement. Why are we reading a config byte for
something we already have in the pci_dev? Why are we now checking
against 0x01 instead of a symbolic constant? Why are we no longer
checking that this is PCI_BASE_CLASS_STORAGE?
> -DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_ATI, PCI_DEVICE_ID_ATI_IXP600_SATA, quirk_sb600_sata);
> -DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_ATI, PCI_DEVICE_ID_ATI_IXP700_SATA, quirk_sb600_sata);
> +DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_ATI, PCI_DEVICE_ID_ATI_IXP600_SATA, quirk_amd_ide_mode);
> +DECLARE_PCI_FIXUP_RESUME(PCI_VENDOR_ID_ATI, PCI_DEVICE_ID_ATI_IXP600_SATA, quirk_amd_ide_mode);
> +DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_ATI, PCI_DEVICE_ID_ATI_IXP700_SATA, quirk_amd_ide_mode);
> +DECLARE_PCI_FIXUP_RESUME(PCI_VENDOR_ID_ATI, PCI_DEVICE_ID_ATI_IXP700_SATA, quirk_amd_ide_mode);
Nothing in the changelog entry suggests why we now need FIXUP_RESUME
entries when we didn't before.
--
Intel are signing my paycheques ... these opinions are still mine
"Bill, look, we understand that you're interested in selling us this
operating system, but compare it to ours. We can't possibly take such
a retrograde step."
> On Thu, Feb 21, 2008 at 03:47:33PM -0800, Greg Kroah-Hartman wrote:
> > +static void __devinit quirk_amd_ide_mode(struct pci_dev *pdev)
> > {
> > - /* set sb600 sata to ahci mode */
> > - if ((pdev->class >> 8) == PCI_CLASS_STORAGE_IDE) {
> > - u8 tmp;
> > + /* set sb600/sb700/sb800 sata to ahci mode */
> > + u8 tmp;
> >
> > + pci_read_config_byte(pdev, PCI_CLASS_DEVICE, &tmp);
> > + if (tmp == 0x01) {
> > pci_read_config_byte(pdev, 0x40, &tmp);
>
> This seems like a dis-improvement. Why are we reading a
> config byte for something we already have in the pci_dev?
> Why are we now checking against 0x01 instead of a symbolic
> constant? Why are we no longer checking that this is
> PCI_BASE_CLASS_STORAGE?
It is a quirk. In pci_ids.h did have PCI_CLASS_STORAGE_IDE
and PCI_BASE_CLASS_STORAGE, these can not represent
the right situation we want to check. 0x01 represents
PCI_CLASS_STORAGE_IDE last 2 bit. Also because it
is a quirk, I do not think we need to change pci_ids.h. So 0x01
used.
> > -DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_ATI,
> > PCI_DEVICE_ID_ATI_IXP600_SATA, quirk_sb600_sata);
> > -DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_ATI,
> > PCI_DEVICE_ID_ATI_IXP700_SATA, quirk_sb600_sata);
> > +DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_ATI,
> > +PCI_DEVICE_ID_ATI_IXP600_SATA, quirk_amd_ide_mode);
> > +DECLARE_PCI_FIXUP_RESUME(PCI_VENDOR_ID_ATI,
> > +PCI_DEVICE_ID_ATI_IXP600_SATA, quirk_amd_ide_mode);
> > +DECLARE_PCI_FIXUP_HEADER(PCI_VENDOR_ID_ATI,
> > +PCI_DEVICE_ID_ATI_IXP700_SATA, quirk_amd_ide_mode);
> > +DECLARE_PCI_FIXUP_RESUME(PCI_VENDOR_ID_ATI,
> > +PCI_DEVICE_ID_ATI_IXP700_SATA, quirk_amd_ide_mode);
>
> Nothing in the changelog entry suggests why we now need
> FIXUP_RESUME entries when we didn't before.
>
PCI configuration space will be changed by BIOS and then in pci
init and restore. So resume also needed.
> --
> Intel are signing my paycheques ... these opinions are still
> mine "Bill, look, we understand that you're interested in
> selling us this operating system, but compare it to ours. We
> can't possibly take such a retrograde step."
>
>
On Fri, Feb 22, 2008 at 01:49:20PM +0800, Cai, Crane wrote:
> > On Thu, Feb 21, 2008 at 03:47:33PM -0800, Greg Kroah-Hartman wrote:
> > > +static void __devinit quirk_amd_ide_mode(struct pci_dev *pdev)
> > > {
> > > - /* set sb600 sata to ahci mode */
> > > - if ((pdev->class >> 8) == PCI_CLASS_STORAGE_IDE) {
> > > - u8 tmp;
> > > + /* set sb600/sb700/sb800 sata to ahci mode */
> > > + u8 tmp;
> > >
> > > + pci_read_config_byte(pdev, PCI_CLASS_DEVICE, &tmp);
> > > + if (tmp == 0x01) {
> > > pci_read_config_byte(pdev, 0x40, &tmp);
> >
> > This seems like a dis-improvement. Why are we reading a
> > config byte for something we already have in the pci_dev?
> > Why are we now checking against 0x01 instead of a symbolic
> > constant? Why are we no longer checking that this is
> > PCI_BASE_CLASS_STORAGE?
> It is a quirk. In pci_ids.h did have PCI_CLASS_STORAGE_IDE
> and PCI_BASE_CLASS_STORAGE, these can not represent
> the right situation we want to check. 0x01 represents
> PCI_CLASS_STORAGE_IDE last 2 bit. Also because it
> is a quirk, I do not think we need to change pci_ids.h. So 0x01
> used.
You haven't explained what is wrong with the original code:
if ((pdev->class >> 8) == PCI_CLASS_STORAGE_IDE) {
> > Nothing in the changelog entry suggests why we now need
> > FIXUP_RESUME entries when we didn't before.
> >
> PCI configuration space will be changed by BIOS and then in pci
> init and restore. So resume also needed.
That information needed to be in the changelog.
--
Intel are signing my paycheques ... these opinions are still mine
"Bill, look, we understand that you're interested in selling us this
operating system, but compare it to ours. We can't possibly take such
a retrograde step."
> On Fri, Feb 22, 2008 at 01:49:20PM +0800, Cai, Crane wrote:
> > > On Thu, Feb 21, 2008 at 03:47:33PM -0800, Greg
> Kroah-Hartman wrote:
> > > > +static void __devinit quirk_amd_ide_mode(struct pci_dev *pdev)
> > > > {
> > > > - /* set sb600 sata to ahci mode */
> > > > - if ((pdev->class >> 8) == PCI_CLASS_STORAGE_IDE) {
> > > > - u8 tmp;
> > > > + /* set sb600/sb700/sb800 sata to ahci mode */
> > > > + u8 tmp;
> > > >
> > > > + pci_read_config_byte(pdev, PCI_CLASS_DEVICE, &tmp);
> > > > + if (tmp == 0x01) {
> > > > pci_read_config_byte(pdev, 0x40, &tmp);
> > >
> > > This seems like a dis-improvement. Why are we reading a
> config byte
> > > for something we already have in the pci_dev?
> > > Why are we now checking against 0x01 instead of a
> symbolic constant?
> > > Why are we no longer checking that this is PCI_BASE_CLASS_STORAGE?
> > It is a quirk. In pci_ids.h did have PCI_CLASS_STORAGE_IDE and
> > PCI_BASE_CLASS_STORAGE, these can not represent the right
> situation we
> > want to check. 0x01 represents PCI_CLASS_STORAGE_IDE last 2
> bit. Also
> > because it is a quirk, I do not think we need to change
> pci_ids.h. So
> > 0x01 used.
>
> You haven't explained what is wrong with the original code:
>
> if ((pdev->class >> 8) == PCI_CLASS_STORAGE_IDE) {
>
When resume, this pdev->class is quirked, however BIOS has modified pci configuration too. Inconsistance occurs.
> > > Nothing in the changelog entry suggests why we now need
> FIXUP_RESUME
> > > entries when we didn't before.
> > >
> > PCI configuration space will be changed by BIOS and then in
> pci init
> > and restore. So resume also needed.
>
> That information needed to be in the changelog.
This info, is a normal info. If maintainer need us to added in source code. I preferred too.
> --
> Intel are signing my paycheques ... these opinions are still
> mine "Bill, look, we understand that you're interested in
> selling us this operating system, but compare it to ours. We
> can't possibly take such a retrograde step."
>
>
>
On Mon, Feb 25, 2008 at 09:43:59AM +0800, Cai, Crane wrote:
> > On Fri, Feb 22, 2008 at 01:49:20PM +0800, Cai, Crane wrote:
> > > > On Thu, Feb 21, 2008 at 03:47:33PM -0800, Greg
> > Kroah-Hartman wrote:
> > > > > +static void __devinit quirk_amd_ide_mode(struct pci_dev *pdev)
> > > > > {
> > > > > - /* set sb600 sata to ahci mode */
> > > > > - if ((pdev->class >> 8) == PCI_CLASS_STORAGE_IDE) {
> > > > > - u8 tmp;
> > > > > + /* set sb600/sb700/sb800 sata to ahci mode */
> > > > > + u8 tmp;
> > > > >
> > > > > + pci_read_config_byte(pdev, PCI_CLASS_DEVICE, &tmp);
> > > > > + if (tmp == 0x01) {
> > > > > pci_read_config_byte(pdev, 0x40, &tmp);
> > > >
> > > > This seems like a dis-improvement. Why are we reading a
> > config byte
> > > > for something we already have in the pci_dev?
> > > > Why are we now checking against 0x01 instead of a
> > symbolic constant?
> > > > Why are we no longer checking that this is PCI_BASE_CLASS_STORAGE?
> > > It is a quirk. In pci_ids.h did have PCI_CLASS_STORAGE_IDE and
> > > PCI_BASE_CLASS_STORAGE, these can not represent the right
> > situation we
> > > want to check. 0x01 represents PCI_CLASS_STORAGE_IDE last 2
> > bit. Also
> > > because it is a quirk, I do not think we need to change
> > pci_ids.h. So
> > > 0x01 used.
> >
> > You haven't explained what is wrong with the original code:
> >
> > if ((pdev->class >> 8) == PCI_CLASS_STORAGE_IDE) {
> >
>
> When resume, this pdev->class is quirked, however BIOS has
> modified pci configuration too. Inconsistance occurs.
Can you update pdev->class from the quirk?
It would be consistent then.
That would leave the code as-is except it's re-reading the field
from config space.
hth,
grant
> > > > Nothing in the changelog entry suggests why we now need
> > FIXUP_RESUME
> > > > entries when we didn't before.
> > > >
> > > PCI configuration space will be changed by BIOS and then in
> > pci init
> > > and restore. So resume also needed.
> >
> > That information needed to be in the changelog.
>
> This info, is a normal info. If maintainer need us to added in source code. I preferred too.
> > --
> > Intel are signing my paycheques ... these opinions are still
> > mine "Bill, look, we understand that you're interested in
> > selling us this operating system, but compare it to ours. We
> > can't possibly take such a retrograde step."
> >
> >
> >
> On Mon, Feb 25, 2008 at 09:43:59AM +0800, Cai, Crane wrote:
> > > On Fri, Feb 22, 2008 at 01:49:20PM +0800, Cai, Crane wrote:
> > > > > On Thu, Feb 21, 2008 at 03:47:33PM -0800, Greg
> > > Kroah-Hartman wrote:
> > > > > > +static void __devinit quirk_amd_ide_mode(struct pci_dev
> > > > > > +*pdev)
> > > > > > {
> > > > > > - /* set sb600 sata to ahci mode */
> > > > > > - if ((pdev->class >> 8) == PCI_CLASS_STORAGE_IDE) {
> > > > > > - u8 tmp;
> > > > > > + /* set sb600/sb700/sb800 sata to ahci mode */
> > > > > > + u8 tmp;
> > > > > >
> > > > > > + pci_read_config_byte(pdev, PCI_CLASS_DEVICE, &tmp);
> > > > > > + if (tmp == 0x01) {
> > > > > > pci_read_config_byte(pdev, 0x40, &tmp);
> > > > >
> > > > > This seems like a dis-improvement. Why are we reading a
> > > config byte
> > > > > for something we already have in the pci_dev?
> > > > > Why are we now checking against 0x01 instead of a
> > > symbolic constant?
> > > > > Why are we no longer checking that this is
> PCI_BASE_CLASS_STORAGE?
> > > > It is a quirk. In pci_ids.h did have PCI_CLASS_STORAGE_IDE and
> > > > PCI_BASE_CLASS_STORAGE, these can not represent the right
> > > situation we
> > > > want to check. 0x01 represents PCI_CLASS_STORAGE_IDE last 2
> > > bit. Also
> > > > because it is a quirk, I do not think we need to change
> > > pci_ids.h. So
> > > > 0x01 used.
> > >
> > > You haven't explained what is wrong with the original code:
> > >
> > > if ((pdev->class >> 8) == PCI_CLASS_STORAGE_IDE) {
> > >
> >
> > When resume, this pdev->class is quirked, however BIOS has modified
> > pci configuration too. Inconsistance occurs.
>
> Can you update pdev->class from the quirk?
I think we can, there are many places where pdev->class is changed in quirks.c.
> It would be consistent then.
> That would leave the code as-is except it's re-reading the
> field from config space.
>
> hth,
> grant
> > > > > Nothing in the changelog entry suggests why we now need
> > > FIXUP_RESUME
> > > > > entries when we didn't before.
> > > > >
> > > > PCI configuration space will be changed by BIOS and then in
> > > pci init
> > > > and restore. So resume also needed.
> > >
> > > That information needed to be in the changelog.
> >
> > This info, is a normal info. If maintainer need us to added
> in source code. I preferred too.
> > > --
> > > Intel are signing my paycheques ... these opinions are still mine
> > > "Bill, look, we understand that you're interested in
> selling us this
> > > operating system, but compare it to ours. We can't possibly take
> > > such a retrograde step."
> > >
> > >
> > >
>
>
>