2006-05-27 05:55:35

by Jeff Garzik

Subject: [PATCH] libata resume improvements

This is an example of how one might make sure the ATA bus reaches
bus-idle, before attempting to talk to it.

It compiles, but is completely untested...

Other resume improvements should work at the pci_driver::resume level as
this does, _not_ the SCSI->ATA->resume device level.


diff --git a/drivers/scsi/ata_piix.c b/drivers/scsi/ata_piix.c
index 6dc8814..949d496 100644
--- a/drivers/scsi/ata_piix.c
+++ b/drivers/scsi/ata_piix.c
@@ -201,7 +201,7 @@ static struct pci_driver piix_pci_driver
.probe = piix_init_one,
.remove = ata_pci_remove_one,
.suspend = ata_pci_device_suspend,
- .resume = ata_pci_device_resume,
+ .resume = pata_pci_device_resume,
};

static struct scsi_host_template piix_sht = {
diff --git a/drivers/scsi/libata-core.c b/drivers/scsi/libata-core.c
index fa476e7..fe42a75 100644
--- a/drivers/scsi/libata-core.c
+++ b/drivers/scsi/libata-core.c
@@ -4852,6 +4852,28 @@ int ata_pci_device_resume(struct pci_dev
pci_set_master(pdev);
return 0;
}
+
+int pata_pci_device_resume(struct pci_dev *pdev)
+{
+ struct device *dev = pci_dev_to_dev(pdev);
+ struct ata_host_set *host_set = dev_get_drvdata(dev);
+ struct ata_port *ap;
+ int i;
+
+ ata_pci_device_resume(pdev);
+
+ msleep(400);
+
+ for (i = 0; i < host_set->n_ports; i++) {
+ ap = host_set->ports[i];
+
+ if (ata_busy_sleep(ap, ATA_TMOUT_BOOT_QUICK, ATA_TMOUT_BOOT))
+ printk(KERN_ERR "ata%u: failed to resume\n", ap->id);
+ }
+
+ return 0;
+}
+
#endif /* CONFIG_PCI */


@@ -4970,6 +4992,7 @@ EXPORT_SYMBOL_GPL(ata_pci_init_one);
EXPORT_SYMBOL_GPL(ata_pci_remove_one);
EXPORT_SYMBOL_GPL(ata_pci_device_suspend);
EXPORT_SYMBOL_GPL(ata_pci_device_resume);
+EXPORT_SYMBOL_GPL(pata_pci_device_resume);
EXPORT_SYMBOL_GPL(ata_pci_default_filter);
EXPORT_SYMBOL_GPL(ata_pci_clear_simplex);
#endif /* CONFIG_PCI */
diff --git a/include/linux/libata.h b/include/linux/libata.h
index b80d2e7..cad531a 100644
--- a/include/linux/libata.h
+++ b/include/linux/libata.h
@@ -516,6 +516,7 @@ extern int ata_pci_init_one (struct pci_
extern void ata_pci_remove_one (struct pci_dev *pdev);
extern int ata_pci_device_suspend(struct pci_dev *pdev, pm_message_t state);
extern int ata_pci_device_resume(struct pci_dev *pdev);
+extern int pata_pci_device_resume(struct pci_dev *pdev);
extern int ata_pci_clear_simplex(struct pci_dev *pdev);
#endif /* CONFIG_PCI */
extern int ata_device_add(const struct ata_probe_ent *ent);

2006-05-27 20:10:30

by Mark Lord

Subject: Re: [PATCH] libata resume improvements

Jeff Garzik wrote:
> This is an example of how one might make sure the ATA bus reaches
> bus-idle, before attempting to talk to it.
>
> It compiles, but is completely untested...
>
> Other resume improvements should work at the pci_driver::resume level as
> this does, _not_ the SCSI->ATA->resume device level.
>
>
> diff --git a/drivers/scsi/ata_piix.c b/drivers/scsi/ata_piix.c
...

Sorry to say, but no user-visible difference from previous attempts.
Still fails the same way, after a 30-ish second timeout on resume.

Cheers

2006-05-27 20:11:55

by Mark Lord

Subject: Re: [PATCH] libata resume improvements

Oops.. replied to the wrong thread, but reply is still valid :)

2006-05-27 20:12:33

by Jeff Garzik

Subject: Re: [PATCH] libata resume improvements

Mark Lord wrote:
> Jeff Garzik wrote:
>> This is an example of how one might make sure the ATA bus reaches
>> bus-idle, before attempting to talk to it.
>>
>> It compiles, but is completely untested...
>>
>> Other resume improvements should work at the pci_driver::resume level as
>> this does, _not_ the SCSI->ATA->resume device level.
>>
>>
>> diff --git a/drivers/scsi/ata_piix.c b/drivers/scsi/ata_piix.c
> ...
>
> Sorry to say, but no user-visible difference from previous attempts.
> Still fails the same way, after a 30-ish second timeout on resume.

Thanks again for testing, though I think more recent of the flurry of
patches invalidated this patch :/ Nonetheless, its useful as a data point.

Jeff