Re: [PATCH] Re: 2.6.17-rc5-git1: regression: resume from suspend(RAM) fails: libata issue

[Jens Axboe <axboe@suse.de>]

On Sat, May 27 2006, Linus Torvalds wrote:
> 
> 
> On Sun, 28 May 2006, Jens Axboe wrote:
> > 
> > Works for me too, I changed the delay to match the 200000 above just to
> > be on the safe side.
> 
> That's insane. That was _not_ what I expected. That wasn't actually 
> supposed to fix anything at all for somebody who had problems.

I realize that, but that is what happens. And I don't think it's a
fluke, it's pretty much 100% reliably failed for me in the later 2.6.17
kernels and this one did 5 resume cycles without even complaining.

> Maybe you didn't remove the half-second delay (which is why it still works 
> for you)? I'd like to see what bits are set when it _isn't_ there, to see 
> if there's something we can trigger on..

The 500msec delay is gone, I can attest to that. This is a straight diff
of what I booted and which apparently works:

diff --git a/drivers/scsi/libata-core.c b/drivers/scsi/libata-core.c
index fa476e7..526d1e1 100644
--- a/drivers/scsi/libata-core.c
+++ b/drivers/scsi/libata-core.c
@@ -4297,6 +4297,13 @@ static int ata_start_drive(struct ata_po
 int ata_device_resume(struct ata_port *ap, struct ata_device *dev)
 {
 	if (ap->flags & ATA_FLAG_SUSPENDED) {
+#if 0
+		ata_busy_wait(ap, ATA_BUSY, 200000);
+#else
+		u8 stat = ata_wait_idle(ap);
+		if (!ata_ok(stat))
+			printk("ATA status %02x\n", stat);
+#endif
 		ap->flags &= ~ATA_FLAG_SUSPENDED;
 		ata_set_mode(ap);
 	}
@@ -4850,6 +4857,7 @@ int ata_pci_device_resume(struct pci_dev
 	pci_restore_state(pdev);
 	pci_enable_device(pdev);
 	pci_set_master(pdev);
+	//msleep(500);
 	return 0;
 }
 #endif /* CONFIG_PCI */
diff --git a/include/linux/libata.h b/include/linux/libata.h
index b80d2e7..80fb1f9 100644
--- a/include/linux/libata.h
+++ b/include/linux/libata.h
@@ -742,7 +742,7 @@ static inline u8 ata_busy_wait(struct at
 
 static inline u8 ata_wait_idle(struct ata_port *ap)
 {
-	u8 status = ata_busy_wait(ap, ATA_BUSY | ATA_DRQ, 1000);
+	u8 status = ata_busy_wait(ap, ATA_BUSY | ATA_DRQ, 200000);
 
 	if (status & (ATA_BUSY | ATA_DRQ)) {
 		unsigned long l = ap->ioaddr.status_addr;

-- 
Jens Axboe