Re: [GIT PULL] Ext3 latency fixes

[Linus Torvalds <torvalds@linux-foundation.org>]

On Mon, 6 Apr 2009, Jeff Garzik wrote:

> Arjan van de Ven wrote:
> > On Sun, 05 Apr 2009 16:57:21 -0400
> > Jeff Garzik <jeff@garzik.org> wrote:
> > > We set it in libata-scsi.c:ata_scsi_dev_config() based on
> > > ata_id_is_ssd()
> > > 
> > > That hueristic probably assumes Intel SSDs or something :/
> > 
> > you mean the "rpm" set to '1' ?
> > I was pretty sure that that was industry standard...
> 
> A -new- industry standard.  You can certainly create a compliant SSD while
> only conforming to ATA-7, for example.  Some older IDE flash devices pretend
> they are normal hard drives in almost every respect, too.

Something like this might be a good idea. 

I've seen several SSD's that do _not_ do that whole RPM == 1 thing, but 
they have "SSD" in their names. 

I forget how the ID is stored (I have this memory of it being big-endian 
16-bit words or something crazy like that?), but aside from fixing up that 
kind of crazyness, maybe something like this is worth it?

And making it non-inline, of course. And maybe it should use 'strstr()' 
instead of checking whether the name ends in 'SSD'. You get the idea..

		Linus

---
--- a/include/linux/ata.h
+++ b/include/linux/ata.h
@@ -769,7 +769,14 @@ static inline int ata_id_is_cfa(const u16 *id)
 
 static inline int ata_id_is_ssd(const u16 *id)
 {
-	return id[ATA_ID_ROT_SPEED] == 0x01;
+	int len;
+	const char *model;
+
+	if (id[ATA_ID_ROT_SPEED] == 0x01)
+		return 1;
+	model = (const char *)&id[ATA_ID_PROD];
+	len = strnlen(model, ATA_ID_PROD_LEN);
+	return len > 3 && !memcmp(model+len-3, "SSD", 3);
 }
 
 static inline int ata_drive_40wire(const u16 *dev_id)