Skip to contentSkip to conversation

Force prefetch for some devices

PATCH In review Jan Evert van Grootheest 4 participants 8 messages Feb 18, 2008 – Jun 27, 2008 libata

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Force prefetch for some devices

Prefetch needs to be set for some ide devices to work when connected to a ht6560b interface. This was not always done properly, causing a system with a HD and CD on the primary interface to not work properly. Or, in effect, hang hard.

This patch forces prefetch on devices before checking whether it is necessary to change the settings in the interface

This patch should also be applied to 2.4. I don't currently have a 2.4 tree around.

(also change my email address)

drivers/ide/legacy/ht6560b.c +16 -7
@@ -23,18 +23,21 @@
  *                      "Prefetch" mode bit OFF for ide disks and
  *                      ON for anything else.
  *
+ *  Version 0.08        Need to force prefetch for CDs and other non-disk
+ *                      devices. (not sure which devices exactly need
+ *                      prefetch)
  *
  *  HT-6560B EIDE-controller support
  *  To activate controller support use kernel parameter "ide0=ht6560b".
  *  Use hdparm utility to enable PIO mode support.
  *
  *  Author:    Mikko Ala-Fossi            <[email protected]>
- *             Jan Evert van Grootheest   <[email protected]>
+ *             Jan Evert van Grootheest   <[email protected]>
  *
  *  Try:  http://www.maf.iki.fi/~maf/ht6560b/
  */
 
-#define HT6560B_VERSION "v0.07"
+#define HT6560B_VERSION "v0.08"
 
 #include <linux/module.h>
 #include <linux/types.h>
@@ -132,15 +135,20 @@
 	u8 select, timing;
 	
 	local_irq_save(flags);
-	
+
 	select = HT_CONFIG(drive);
 	timing = HT_TIMING(drive);
-	
+
+	/*
+	 * Need to enforce prefetch sometimes because otherwise
+	 * it'll hang (hard).
+	 */
+	if (drive->media != ide_disk || !drive->present)
+		select |= HT_PREFETCH_MODE;
+
 	if (select != current_select || timing != current_timing) {
 		current_select = select;
 		current_timing = timing;
-		if (drive->media != ide_disk || !drive->present)
-			select |= HT_PREFETCH_MODE;
 		(void)inb(HT_CONFIG_PORT);
 		(void)inb(HT_CONFIG_PORT);
 		(void)inb(HT_CONFIG_PORT);
@@ -190,11 +198,12 @@
 	outb(HT_TIMING_DEFAULT, 0x1f6);  /* IDE_SELECT_REG */
 	(void) inb(0x1f7);               /* IDE_STATUS_REG */
 	
-	printk("\nht6560b " HT6560B_VERSION
+	printk("ht6560b " HT6560B_VERSION
 	       ": chipset detected and initialized"
 #ifdef DEBUG
 	       " with debug enabled"
 #endif
+	       "\n"
 		);
 	return 1;
 }
#1 lore raw

Hello.

Quoted 1 line · Jan Evert van Grootheest

Prefetch needs to be set for some ide devices to work when connected to

Prefetch bit seems to mean 1=disable, 0=enable despite what the datasheet
says...

Quoted 7 lines

a ht6560b interface. This was not always done properly, causing a system with a HD and CD on the primary interface to not work properly. Or, in effect, hang hard.

This patch forces prefetch on devices before checking whether it is necessary to change the settings in the interface

Quoted 6 lines

This patch should also be applied to 2.4. I don't currently have a 2.4 tree around.

(also change my email address)

Please send the patch to the current 2.4 maintainer, after applying it to
the 2.4 tree which you can download on kernel.org.

drivers/ide/legacy/ht6560b.c line 29

Quoted 12 lines

------------------------------------------------------------------------

drivers/ide/legacy/ht6560b.c +3 -0
@@ -23,18 +23,21 @@
  *                      "Prefetch" mode bit OFF for ide disks and
  *                      ON for anything else.
  *
+ *  Version 0.08        Need to force prefetch for CDs and other non-disk
+ *                      devices. (not sure which devices exactly need
+ *                      prefetch)
  *

Hard disks only ai guess, i.e. non-ATAPI device.

MBR, Sergei

#2 lore raw
In reply to Jan Evert van Grootheest
Quoted 7 lines · On Mon, 18 Feb 2008 13:50:02 +0100 Jan Evert van Grootheest…

Prefetch needs to be set for some ide devices to work when connected to a ht6560b interface. This was not always done properly, causing a system with a HD and CD on the primary interface to not work properly. Or, in effect, hang hard.

This patch forces prefetch on devices before checking whether it is necessary to change the settings in the interface

Interesting, that is backwards to the docs and not something I've seen. If you are correct then pata_legacy also wants tweaking .

Alan

#3 lore raw

Alan,

Quoted 9 lines · Alan Cox
Quoted 5 lines · On Mon, 18 Feb 2008 13:50:02 +0100 Jan Evert van Grootheest…

Prefetch needs to be set for some ide devices to work when connected to a ht6560b interface. This was not always done properly, causing a system with a HD and CD on the primary interface to not work properly. Or, in effect, hang hard.

Quoted 6 lines
Quoted 3 lines

This patch forces prefetch on devices before checking whether it is necessary to change the settings in the interface

Interesting, that is backwards to the docs and not something I've seen.

Hum, I may have an incorrect understanding of the prefetch bit or misread the code. But for sure that patch makes my 486 boot (HD as primary master and CD as primary slave).

From testing I know that the HD works with and without prefetch (it's 1.15M/s without prefetch, ~950k/s with). And the boot hangs (as in powerbutton required) if prefetch is not set for the CD. For reference, the generic IDE driver is ~850k/s.

Quoted 4 lines

If you are correct then pata_legacy also wants tweaking .

Alan

I think, but didn't look real good nor test yet, that pata_legacy also needs something comparable to select_proc? Or is that implemented in some generic manner? Depending on the day job, I could get to testing that somewhere this week.

-- Jan Evert

#4 lore raw
Quoted 3 lines

I think, but didn't look real good nor test yet, that pata_legacy also needs something comparable to select_proc? Or is that implemented in some generic manner?

It sets it in the timing routine:

	pata_legacy:ht6560b_set_piomode
#5 lore raw
In reply to Alan Cox

Alan, Bart,

It's been a while...

Quoted 9 lines · Alan Cox
Quoted 5 lines · On Mon, 18 Feb 2008 13:50:02 +0100 Jan Evert van Grootheest…

Prefetch needs to be set for some ide devices to work when connected to a ht6560b interface. This was not always done properly, causing a system with a HD and CD on the primary interface to not work properly. Or, in effect, hang hard.

Quoted 7 lines
Quoted 3 lines

This patch forces prefetch on devices before checking whether it is necessary to change the settings in the interface

Interesting, that is backwards to the docs and not something I've seen. If you are correct then pata_legacy also wants tweaking .

I'm testing this... but lack some knowledge.
Is there some generic way to figure out wheter a device supports it?
What devices normally support prefetching? Disks only?
I already found out that CDs do not. But tapes? Floppies? What else can be there that might support prefetching?

Thanks,
Jan Evert

#6 lore raw
Quoted 2 lines

Is there some generic way to figure out wheter a device supports it? What devices normally support prefetching? Disks only?

It depends on the controller not the device - generally older controller support it only for disk devices.

Alan

#7 lore raw
In reply to Jan Evert van Grootheest
Quoted 8 lines · Jan Evert van Grootheest

Prefetch needs to be set for some ide devices to work when connected to a ht6560b interface. This was not always done properly, causing a system with a HD and CD on the primary interface to not work properly. Or, in effect, hang hard.

This patch forces prefetch on devices before checking whether it is necessary to change the settings in the interface

Quoted 6 lines

This patch should also be applied to 2.4. I don't currently have a 2.4 tree around.

(also change my email address)

I applied it as it is but could I ask you to follow up on the following (rather minor) issues:

- According to observations from your testing and comments from Sergei/Alan it really seems that prefetch bit has reversed meaning so the driver should be updated to reflect it.

- Could the old changelog be removed while we are at it? The detailed changelog is kept in the kernel git tree nowadays.

Thanks,
Bart