Additions to libata patch archive

Additions to libata patch archive

[Jeff Garzik]

I just added a few more patches to the libata patch archive.  The patch 
archive is a collection of libata patches that aren't being actively 
maintained, but could potentially be useful as a reference for a better 
patch in the future.


http://www.kernel.org/pub/linux/kernel/people/jgarzik/libata/archive/2.6.12-rc5-git8-ioctl-get-identity.patch.bz2
This patch adds support for the HDIO_GET_IDENTITY ioctl.  Bart (IDE 
maintainer) raised some valid objections -- objections that I keep 
forgetting once months pass -- and so the patch just sat.


http://www.kernel.org/pub/linux/kernel/people/jgarzik/libata/archive/2.6.13-vsc-mwi.patch.bz2
More properly configure Vitesse 7174 / Intel 31244 SATA controllers' 
MSI, DMA burst, and cacheline size settings.  Jeremy Higdon tested this 
patch and didn't see any problems, but since the current driver has been 
working, I didn't see any need to rush this patch in.  Maybe someone 
with hardware will become interested in it, as it does program the 
hardware more correctly.


http://www.kernel.org/pub/linux/kernel/people/jgarzik/libata/archive/2.6.14-rc3-sil-lbt.patch.bz2
This patch turns on the Silicon Image 311x "Large Block Transfer" 
feature, which eliminates legacy PCI IDE DMA limitations such as 64k 
boundary and 64k max DMA segment size.  This needs a lot of stress 
testing, but really should get merged as soon as enough people verify 
that it works for them.


http://www.kernel.org/pub/linux/kernel/people/jgarzik/libata/archive/2.6.14-rc3-sil24-original.patch.bz2
Original Silicon Image 3124 driver contributed by Silicon Image.  Since 
there is a newer driver, this is merely provided as a reference.

Re: Additions to libata patch archive

[Bryan O'Sullivan]

On Wed, 2005-10-05 at 00:13 -0400, Jeff Garzik wrote:

> http://www.kernel.org/pub/linux/kernel/people/jgarzik/libata/archive/2.6.14-rc3-sil24-original.patch.bz2
> Original Silicon Image 3124 driver contributed by Silicon Image.  Since 
> there is a newer driver, this is merely provided as a reference.

In case anyone should care, the original SiImage driver oopses during
boot with recent kernels.

The reason I know this is that I just tried booting 2.6.14-rc3 with the
0.10 (SiImage) driver.  It appears to contain port multiplier support,
which I need.  Tejun's rewrite doesn't, and SiImage wants me to sign an
NDA to get a more recent driver of theirs that has port multiplier
support.

	<b

Re: Additions to libata patch archive

[Edward Falk]

> In case anyone should care, the original SiImage driver oopses during
> boot with recent kernels.
> 
> The reason I know this is that I just tried booting 2.6.14-rc3 with the
> 0.10 (SiImage) driver.  It appears to contain port multiplier support,
> which I need.  Tejun's rewrite doesn't, and SiImage wants me to sign an
> NDA to get a more recent driver of theirs that has port multiplier
> support.

The original vendor-supplied driver did not support port multipliers, 
but it had the hooks that implied they were thinking about it.  The 
original driver had a number of problems, mostly dealing with lack of 
error recovery as you have just discovered.

Tejun's driver is much better, and I have sent him some patches that 
will make it quite stable.


There are two approaches to port multiplier support -- one is to add it 
to the driver (a real kludge) or to add it to libata.  Kludging it into 
the driver is a quick-and-dirty solution, while adding it to libata will 
take more time.  I'm trying to find the time to do the latter.

	-ed falk