Hi, Jeff.

 My posted patch is broken anyway.  I was indexing sil_port_info w/
bit masks.  :-(

On Mon, Aug 22, 2005 at 03:36:57PM -0400, Jeff Garzik wrote:

Tejun Heo wrote:

quoted

Hello, Jeff & Bartlomiej & all.

quoted

What about fixing 3114 for a start? :-)

Yeap, why not?  Here it is.
--

This patch adds chipset mask to sil_blacklist such that quirks can be
applied selectively depending on chipset.  And add sil_3112_m15w
chipset which is identical to sil_3112 except that mod15write quirk is
applied.  As we don't know which PCI IDs are affected yet, I've
changed all 3112's to sil_3112_m15w.  Later when we know which PCI IDs
are affected, we can just change them to sil_3112 later.


Signed-off-by: Tejun Heo <redacted>

Thanks for tackling this.  I would approach it in a different manner:

1) add to the enum at the top,
+	SIL_FLAG_M15W		= (1 << 30),

2) do
        sil_3112                = 0,
-       sil_3114                = 1,
+	sil_3112_m15w		= 1,
+	sil_3114		= 2,


3) in sil_pci_tbl[], change all the sil_3112 entries to sil_3112_m15w, 
EXCEPT for PCI ID 0x1095, 0x3512, which should remain sil_3112.



4) in sil_port_info[], duplicate sil_3112 entry to new entry 
sil_3112_m15w.  make sure order in sil_port_info[] array matches the 
values in step #2.


5) add SIL_FLAG_M15W to the list of host_flags in the new sil_3112_m15w 
entry created in step #4.

 Can we use host_flags for that?  It seems that it's for ATA_FLAG_*
flags, and we currently don't explicitly reserve any bits from
ATA_FLAG_*.

6) in sil_dev_config(), change
       if (quirks & SIL_QUIRK_MOD15WRITE) {
test to also test (ap->flags & SIL_FLAG_M15W)


Check over these steps to see if I missing anything...  at the very 
least, I have illustrated the approach that should be taken.

	Jeff

 Thanks.

-- 
tejun