[SATA] libata-dev queue updated, status report

libata-dev-2.6, the development queue for the Linux SATA driver libata, 
has been updated.  See attachment for BK info, patch URL, and full list 
of changes.

Here is a quick summary of the goodies lurking in libata-dev, and also 
the Reason Why It's Not Upstream (RWINU):


1) ATA passthru.  Arbitrary ATA command execution.  Provides SMART and 
hdparm support, among other things.

RWINU:  SCSI T10 committee will not standardize the opcodes for this 
until March 2005.  The "set transfer mode" command must be 
special-cased, since it requires additional per-controller work to 
reconfigure DMA timings and such.


2) Turn on ATAPI.

RWINU:  Needs some more testing, and need to resolve discrepancy between 
ATA/ATAPI-4 and ATA/ATAPI-7 for ATAPI DMA host state machine.


3) Detect PATA devices attached to SATA controllers (via a bridge), and 
adjust settings to limit udma mode and maximum number of sectors, due to 
  problems with certain bridge/controller combinations and issues 
reported in the field.

RWINU: Not certain that the hueristic reliably detects a PATA device 
attached to a SATA controller.


4) Cleanup:  Remove 'execute device diagnostic' reset support, making 
code smaller.

RWINU: Need to make sure no out-of-tree drivers are using it.


5) hdparm 'get identity' ioctl support.

RWINU: Bart had some objections that warranted attention (but I forget 
what those objections were, alas).


Additional hardware support
---------------------------

1) New driver:  Promise 2027x (PATA)

RWINU:  Need to plan a good strategy for users who wish to use this 
driver, versus the PATA driver in drivers/ide.  Also applies to other 
PATA drivers people will eventually port to libata.


2) New driver:  ADMA.  Supports several Pacific Digital controllers, 
among others.

RWINU: Never tested on real hardware; needs testing.


3) New driver:  VIA VT6421.

RWINU: Never tested on real hardware; needs testing.


4) Support for PATA port on Promise SATA controllers.

RWINU:  I'm not sure whether the API change necessary for this is the 
correct one.  Need to ponder.  Comments welcome.


5) Support for Promise PDC20619 (a.k.a. TX4000) PATA controller.

RWINU:  Needs additional testing.  Need to rename 'sata_promise' to 
'ata_promise', then create a method of aliasing the old module name, to 
avoid breaking existing setups.  2.6.x has the MODULE_ALIAS() method, 
2.4.x needs a Makefile rule to build an additional module.


6) core libata PATA support.

RWINU:  PATA is working, but needs a few missing pieces filled in.   CHS 
support is one (Albert Lee just started on this, posting the first 
patches to linux-ide last night), DMA blacklist is another (done; 
upstream soon), and a few other details.  Also the issues mentioned in 
"#1 Promise PDC2027x" (above).