updated new EH and hotplug support are about to be posted...

updated new EH and hotplug support are about to be posted...

[Tejun Heo]

Hello, people.

It's been another looooooong week.  Build number of my ATA work tree,
which was created at the end of Feburary, is now over 1040 and all my
SATA plugs are worn and unplug easily.  But, hey, hotplug works now.

I'm about to post the following 9 patchsets composed of 97 patches.
The first two patchsets have been posted previously but I'll repost
them for simplicity's sake.

1. misc-reset-updates			: repost (3 patches)
2. implement-and-use-ata_wait_register	: repost (5 patches)
3. misc-ata_bus_probe-updates		: take 1 (3 patches)
4. fixes-errata-workaround-and-reset-updates : take 3 (13 patches)
5. new-EH-framework			: take 2 (16 patches)
6. new-EH-implementation		: take 2 (14 patches)
7. add-NCQ-support			: take 3 (15 patches)
8. prep-for-hotplug-support		: take 1 (15 patches)
9. add-hotplug-support			: take 1 (13 patches)

All patchsets are against the current upstream[1] and stacks on the
previous ones.  However, patchsets 5 and upwards depend on the
following SCSI updates.

implement-scsi_eh_schedule patchset [2]
fix-scsi_kill_request-busy-count-handling patch [3]

new-EH-framwork and new-EH-implementation patchsets received a lot of
updates (all recorded in the head message of the patchests).  NCQ
patchset is largely unchanged except other than addition of sata_sil24
support.

Hotplug supports both warm and hot unplug/plug.  More details are on
the head message of add-hotplug-support patchset and inidividual patch
descriptions.

I'm planning on making combined patches against release kernels
(stable kernel and maybe -rc?) such that the whole EH/NCQ/hotplug can
be tested more widely and impatient and living-on-the-cutting-edge
people don't have to wait till these make it to mainline.

Also, if you guys have any major objection or concerns with the
current design and implementation.  Please let me know early as I'm
gonna start working on port multiplier support on top of these very
soon.

Albert, sata_sil-new_interrupt_handler patch in the patchset
add-hotplug-support implements a private interrupt handler for
sata_sil but sata_sil still needs to use PIO support in libata.
Currently it's okay as irq-driven path and PIO path are completely
separate, but to merge with irq-pio branch, we need to separate PIO
HSM from the standard interrupt handler such that LLDD's private
interrupt handler can drive HSM.

Thanks!

--
tejun

[1] c2a6585296009379e0f4eff39cdcb108b457ebf2
[2] http://article.gmane.org/gmane.linux.ide/9290
[3] http://article.gmane.org/gmane.linux.ide/9487

Re: updated new EH and hotplug support are about to be posted...

[zhao, forrest]

On Tue, 2006-04-11 at 22:08 +0900, Tejun Heo wrote:
> Hello, people.
> 
> It's been another looooooong week.  Build number of my ATA work tree,
> which was created at the end of Feburary, is now over 1040 and all my
> SATA plugs are worn and unplug easily.  But, hey, hotplug works now.
> 
> I'm about to post the following 9 patchsets composed of 97 patches.
> The first two patchsets have been posted previously but I'll repost
> them for simplicity's sake.
> 
> 1. misc-reset-updates			: repost (3 patches)
> 2. implement-and-use-ata_wait_register	: repost (5 patches)
> 3. misc-ata_bus_probe-updates		: take 1 (3 patches)
> 4. fixes-errata-workaround-and-reset-updates : take 3 (13 patches)
> 5. new-EH-framework			: take 2 (16 patches)
> 6. new-EH-implementation		: take 2 (14 patches)
> 7. add-NCQ-support			: take 3 (15 patches)
> 8. prep-for-hotplug-support		: take 1 (15 patches)
> 9. add-hotplug-support			: take 1 (13 patches)
> 

Tejun,

Where can I find your git tree, which includes all the above
patch sets? We're preparing to test SATA hotplug in our lab.

BTW. Is there a way to check whether SATA controller support
hotplug, such as a capability bit in certain register?
Or we have to first enable PHY status changed interrupt, then
hotplug the SATA disk, then to check if the interrupt is triggered,
and PHY RDY status changes?

I read AHCI 1.1 spec, and didn't find such capability bit.

Thanks,
Forrest

Re: updated new EH and hotplug support are about to be posted...

[Jeff Garzik]

zhao, forrest wrote:
> BTW. Is there a way to check whether SATA controller support
> hotplug, such as a capability bit in certain register?

For AHCI, all controllers support hotplug.

Most SATA controllers support hotplug, across all vendors.

If you're lucky, silicon has a phy interrupt.  If you're somewhat lucky, 
silicon has a PhyRdy state interrupt.  If you're stuck with a BMDMA 
interface with bolted-on SATA PHY registers, hotplug will -probably- work.


> Or we have to first enable PHY status changed interrupt, then
> hotplug the SATA disk, then to check if the interrupt is triggered,
> and PHY RDY status changes?

The interrupt is quite helpful, but ultimately optional.

	Jeff