Re: Loud "pop" coming from hard drive on reboot

[Mark Lord <lkml@rtr.ca>]

Alan Cox wrote:
>
> If you see a synchronize cache succeed and you then see the drive
> shutdown succeed then you know that a sync cache can be faked as ok
> safely. Any other command in between or after and it doesn't get faked
> 
> This seems pretty easy to deal with at command issue.

Yup.  It could be as trivial as this (untested, and not as nice
as Alan's full suggestion).

Signed-off-by:  Mark Lord <mlord@pobox.com>
---
--- linux/include/linux/libata.h.orig	2007-04-18 10:30:25.000000000 -0400
+++ linux/include/linux/libata.h	2007-04-18 10:30:28.000000000 -0400
@@ -499,6 +499,7 @@
 	struct ata_ering	ering;
 	int			spdn_cnt;
 	unsigned int		horkage;	/* List of broken features */
+	int			needs_flush;	/* bool: 1==written since last flush */
 #ifdef CONFIG_SATA_ACPI
 	/* ACPI objects info */
 	acpi_handle obj_handle;
--- linux/drivers/old/libata-scsi.c	2007-04-17 22:11:12.000000000 -0400
+++ linux/drivers/ata/libata-scsi.c	2007-04-18 10:27:29.000000000 -0400
@@ -2749,18 +2749,20 @@
 			return atapi_xlat;
 
 	switch (cmd) {
-	case READ_6:
-	case READ_10:
-	case READ_16:
-
 	case WRITE_6:
 	case WRITE_10:
 	case WRITE_16:
+		dev->needs_flush = 1;
+	case READ_6:
+	case READ_10:
+	case READ_16:
 		return ata_scsi_rw_xlat;
 
 	case SYNCHRONIZE_CACHE:
-		if (ata_try_flush_cache(dev))
+		if (dev->needs_flush && ata_try_flush_cache(dev)) {
 			return ata_scsi_flush_xlat;
+			dev->needs_flush = 0;
+		}
 		break;
 
 	case VERIFY: