Alan Cox [interview] offered an updated libATA PATA (IDE) status report. He summarized, "with the exception of HPA and serialize support its now pretty close to a straight replacement for drivers/ide on x86 systems (and boxes using PCI devices only). There is other stuff that wants improving still like error recovery on CRC, but its getting close."

He provides a list of recent changes then cautions, "please remember that functionality equivalence, and much cleaner code doesn't mean less bugs yet, there is a *lot* of testing and hammering on the code needed before it is production ready for switching." His current patch is for the 2.6.16-rc2 kernel.

From: Alan Cox [email blocked]
To:  linux-ide, [email blocked]
Subject: libATA  PATA status report, new patch
Date:	Mon, 06 Feb 2006 16:46:52 +0000

Main changes this time
- Initial driver code for PCMCIA
- Initial driver code for ISAPnP
- Legacy support including VLB driver bits (mostly to test non PCI
cases)
- Replace IRQ masking awareness in core with a ->data_xfer method
- Use ->data_xfer to move IRQ masking, VLB sync and other chip 
  whackiness back *outside* of the core libata code.
- Simplex DMA is now supported
- Remove PCI quirk junk for IDE stuff (its buggy in places and a mess)
- Teach core PCI code about the IDE legacy mess in the PCI spec
- Merge with other libata changes to 2.6.16-rc2
- ARTOP driver now knows how to init Macintosh cards
- Initial netcell driver bits
- CMD64x support (except 640)
- Initial incomplete Cypress Alpha IDE


With the exception of HPA and serialize support its now pretty close to
a straight replacement for drivers/ide on x86 systems (and boxes using
PCI devices only). There is other stuff that wants improving still like
error recovery on CRC, but its getting close.

Please remember that functionality equivalence, and much cleaner code
doesn't mean less bugs yet, there is a *lot* of testing and hammering on
the code needed before it is production ready for switching.

	http://zeniv.linux.org.uk/~alan/IDE

for 2.6.16-rc2 patches.

Alan



Core features needed for good PATA

-	pre LBA disk support				(In progress elsewhere)
-	HPA management
	Power management				(In progress elsewhere)
-	Hotplug/Warmplug				(In progress elsewhere)
-	Per platform legacy port data/quirks
DONE	Smarter mode filtering hooks
DONE	Legacy simplex support
-	Serialize support
DONE	IRQ masking

Obscurer stuff
-	Mixed mode legacy/native
-	Removable detection
-	EIDE mode timing support
-	"guess the system bus speed"

Please test and report but don't be surprised if it oopses on boot or eats 
your disk. For now please report success and failures so I can get an idea of 
coverage and patterns.

When filing a report include

	lspci -vxxx of board (off any kernel)
	info on all disks (hdparm info off old driver is good)
	40/80pin cable notes

to
 [email blocked]

patches, offers of help and the like also welcome.

Other Notes:
	The core libata changes I am using may or may not end up in the 
upstream code as I work with Jeff Garzik to get them into a form he wants.
Don't rely on them for SATA drivers just the PATA patch work.


Device Status
=============

PCI & Later
===========

ALi:
	Initial driver.
	LBA48 PIO with LBA28 DMA
	No MWDMA

AMD:
	Complete

ARTOP:
	Initial driver
	Added Macintosh bits

ATIIXP:
	Complete, testing work required

CMD640:
	No driver yet. PCI driver planned.

CMD64x:
	Initial port

CS5520:
	Working PIO, no DMA support yet
	Fixed unload crash

CS5530:
	Passed initial tests
	Cannot yet use MWDMA and UDMA devices together

CS5535:
	Complete, needs testing work

Cypress (Alpha specific):
	Initial evaluation only

Delkin:
	Cardbus ATA controller. No driver, out of tree old ide driver.

EFAR: 
	Initial port

Generic:
	Complete, was working, not tested recently. 

HPT343/363:
	Initial port

HPT36x:
	Initial port

HPT37x:
	Initial port. Basic testing done.

HPT3x2n:
	Initial port

IT8172:
	Initial port

IT821x:
	Initial port, not heavily tested yet
	Should be fully functional in passthru
	Smart needs	Command filter
			RAID ioctls
			ATAPI smarts
	Needs a couple of other core changes yet

MPIIX:
	Intel MPIIX mobile chipset. Complete

NS84710:
	Handled by generic driver as with old IDE. Datasheet available
	if you have this relic and want to work on it.

NS84715:
	Being handled by someone else

OLDPIIX:
	Support for original PIIX chipset. Complete.

Opti:
	Initial port, should work to PIO4. Complete.

PIIX:
	Should support all chipsets but PIIX/MPIIX.

Promise:
	Being handled by someone else.

RZ1000:
	Simulation tested only. Complete.

SC1200:
	Untested, may need serialize support in the core

Serverworks:
	Minimal testing.

SGIIOC4:
	Not supported

SIL680:
	Heavily tested.
	Some combinations of PIO drives don't yet work it seems ?

SIS:
	Initial driver
	Basic testing completed

SL82C05:
	Initial port

Triflex:
	Working and tested (but not tested recently)

TRM290:
	Weird pre standardisation DMA controller. Not supported. No plans
	to bother

VIA:
	Added IRQ masking support
	Completed


Pre PCI
=======

ALI14xx:
	Not supported

DTC2278:
	Not supported. No plans

HT6560:
	PIO tuning handled by pata_legacy

ISAPnp:
	Initial driver, needs testing work.

Legacy:
	Now handles primary/secondary non PCI ports

Opti82C611A:
	Handled by legacy driver

PCMCIA:
	Initial driver. Needs a lot of testing yet.

PDC20230-C/20630:
	Autodetection and PIO tuning in pata_legacy.
	No 20630 wacky DMA plans.

PDC4030:
	No support. (old driver for reference but
	seems not worth it due to age and complexity)

QD65xx:
	Test driver to prove non PCI code paths.	

UMC8672:
	Not supported. No plans.

Winbond 83759A:
	Not supported. Use jumpers to set speeds. Data sheet exists.

• Archive of above thread
• KernelTrap interview with Alan Cox