libata: add warning about possible HPA compatibility issue

Document "ignore_hpa" module parameter and add a warning about possible HPA compatibility issue.

This should fix bug #13365:

	http://bugzilla.kernel.org/show_bug.cgi?id=13365

To: Jeff Garzik <[email protected]>

---
If people are fine with the patch I can push it together with ide fixes for 2.6.30-rc8.

 Documentation/kernel-parameters.txt |    6 ++++++
 drivers/ata/libata-core.c           |   15 +++++++++++++--
 2 files changed, 19 insertions(+), 2 deletions(-)

Index: b/Documentation/kernel-parameters.txt
===================================================================
@@ -1101,6 +1101,12 @@ and is between 256 and 4096 characters. 
 			If there are multiple matching configurations changing
 			the same attribute, the last one is used.
 
+	libata.ignore_hpa=
+			[LIBATA] Ignore HPA limit
+			Format: { "0" | "1" }
+			0 -- keep BIOS limits (default)
+			1 -- ignore limits, using full disk
+
 	lmb=debug	[KNL] Enable lmb debug messages.
 
 	load_ramdisk=	[RAM] List of ramdisks to load from floppy

Index: b/drivers/ata/libata-core.c
===================================================================
@@ -1472,6 +1472,16 @@ static int ata_set_max_sectors(struct at
 	return 0;
 }
 
+static inline void ata_hpa_warn(unsigned long long first_sector,
+				unsigned long long last_sector)
+{
+	printk(KERN_WARNING "*** possible compatibility issue:\n"
+		"*** sectors %llu-%llu may be inaccessible\n"
+		"*** if Host Protected Area is used for filesystem data\n"
+		"*** (please refer to \"ignore_hpa\" module parameter)\n",
+		first_sector, last_sector);
+}
+
 /**
  *	ata_hpa_resize		-	Resize a device with an HPA set
  *	@dev: Device to resize

@@ -1521,12 +1531,13 @@ static int ata_hpa_resize(struct ata_dev
 		if (!print_info || native_sectors == sectors)
 			return 0;
 
-		if (native_sectors > sectors)
+		if (native_sectors > sectors) {
 			ata_dev_printk(dev, KERN_INFO,
 				"HPA detected: current %llu, native %llu\n",
 				(unsigned long long)sectors,
 				(unsigned long long)native_sectors);
-		else if (native_sectors < sectors)
+			ata_hpa_warn(sectors, native_sectors - 1);
+		} else if (native_sectors < sectors)
 			ata_dev_printk(dev, KERN_WARNING,
 				"native sectors (%llu) is smaller than "
 				"sectors (%llu)\n",

Please propose better description.

As was discussed before

Tejun has a patch in the queue which exposes the HPA and real sizes to the OS

Libata can support revalidating a disc

Once the HPA/Real size data is there the tools can be tuned for this and a sysfs node added to switch hpa on/off per disc. At that point this can be done by the user space intelligently.

The fact this is a user parameter is itself a problem.

In queue for 2.6.30?

Unless you happen to have partition with tools in HPA area etc.

Also care to post link to the existing user-space tools with added HPA support and documentation how to use them so I can start pointing people reporting problems in this direction?

Yes. It is a problem but we can't reverse the history.

I fully agree w.r.t to proper handling of HPA but we need "something" in the meantime because it is *real* issue resulting in *real* users filling *real* bug-reports (which unfortunately always get assigned to me first so I'm needlessly wasting my time on them).

IOW Please stop downplaying bug #13365 which this patch tries to alleviate.

I'm adding Andrew to cc:...

2.6.30 merge window is closed, as you well kow

The merge window is closed. This is a 2.6.31 discussion and randomly adding bogus printks to the kernel won't help because

- Users have no idea what they mean
- Distros boot with "quiet" so users don't see them anyway

It would be good to get Tejun's patch upstream and also runtime HPA switching.

I'm still not convinced 13365 is an HPA problem entirely - but there are certainly other HPA confusions here and there.

Hello Andrew ;)

Alan