PCI patches for 2.6.24-rc1

Here are a some PCI patches against your 2.6.24-rc1 git tree.

They are a bunch of quirk updates from David Miller, a new config item to help Jeff Garzik start to cleanup the isdn drivers and let him take those patches through his tree, and a few other minor bugfixes.

All of these have been in the -mm tree for a while.

Please pull from:

	master.kernel.org:/pub/scm/linux/kernel/git/gregkh/pci-2.6.git/

The full patches will be sent to the linux-pci mailing list, if anyone wants to see them.

thanks,

greg k-h

-------------

 drivers/isdn/hisax/Kconfig        |   18 ++++++------
 drivers/isdn/hisax/avm_pci.c      |    4 +-
 drivers/isdn/hisax/diva.c         |    6 ++--
 drivers/isdn/hisax/elsa.c         |    4 +-
 drivers/isdn/hisax/gazel.c        |    4 ++-
 drivers/isdn/hisax/niccy.c        |    7 +++--
 drivers/isdn/hisax/sedlbauer.c    |    4 +-
 drivers/net/tg3.c                 |    9 ------
 drivers/pci/Kconfig               |   11 ++++++++
 drivers/pci/hotplug/Kconfig       |    6 ++--
 drivers/pci/hotplug/cpqphp_ctrl.c |   16 +++++------
 drivers/pci/msi.c                 |   18 ++++++++----
 drivers/pci/pci-driver.c          |    5 +--
 drivers/pci/quirks.c              |   51 ++++++++++++++++++++++++++++++++++---
 drivers/pci/search.c              |    9 ++++++
 drivers/scsi/Kconfig              |    2 +-
 drivers/serial/8250_pci.c         |    5 +++-
 include/linux/pci.h               |   14 ++++++++-
 include/linux/pci_ids.h           |    5 +---
 19 files changed, 134 insertions(+), 64 deletions(-)

---------------

Adrian Bunk (2):
PCI: make pci_match_device() static
PCI Hotplug: cpqhp_pushbutton_thread(): remove a pointless if() check

David Miller (5):
PCI: Revert "PCI: disable MSI by default on systems with Serverworks HT1000 chips"
PCI: Add MSI quirk for ServerWorks HT1000 PCIX bridge.
PCI: Add quirk for devices which disable MSI when INTX_DISABLE is set.
PCI: Add MSI INTX_DISABLE quirks for ATI SB700/800 SATA and IXP SB400 USB
PCI: Remove 3 incorrect MSI quirks.

Dirk Hohndel (1):
PCI: pciserial_resume_one ignored return value of pci_enable_device

Jeff Garzik (1):
PCI: Add Kconfig option to disable deprecated pci_find_* API

This one is bogus:

PCI: Revert "PCI: disable MSI by default on systems with Serverworks HT1000 chips"

This reverts commit e3008dedff4bdc96a5f67224cd3d8d12237082a0.

The real bug was an INTX issue in the tg3 ethernet chip, and cured by commit c129d962a66c76964954a98b38586ada82cf9381

It says "cured by commit c129d962a66c76964954a98b38586ada82cf9381", but no such commit exists. What?

Did you mean commit ba698ad4b7e466cbb4a8bde6b9da8080ab06808d, and if so, where did that c129d.. come from?

Linus

Side note: I pulled and pushed out, since I assume the *code* is correct, but please double-check the commit messages for things like this.

gitk makes it really easy, since it will highlight valid commit ID's. It looks lik eyou may have re-ordered the commits or something?

Linus

Wow, that's cool, I never noticed that.

I'm guessing that David is referring to a commit in his tree, not in yours yet.

thanks,

greg k-h

Well, I suspect that David referred to a commit that he just sent by email. Which obviously will have a *different* commit ID once you commit it - so it probably made sense in his tree, but not once he exported it as an email instead of syncing with git natively.

I also suspect that the whole series was re-ordered in email (or by you reading/applying them in a different order). Since now the revert of the "disable MSI" happens *before* the patch that looks like it will fix the issue.

So I think David also probably didn't number his patches.

Linus

From: Linus Torvalds <[email protected]>
Date: Mon, 5 Nov 2007 14:42:15 -0800 (PST)

They were in the correct order, and I did number them.

The commit with the invalid commit ID in question went in a long time ago, into 2.6.23 in fact, and as I explained in another email I rebased by net-2.6 tree before I had sent that patch in, but I forgot to update the commit log in these PCI layer changes.

I was not referencing a commit ID within the PCI patch set itself.

I just did a pull and verified that the commits are in the correct order.

For reference the commit I meant to reference was:

commit 2fbe43f6f631dd7ce19fb1499d6164a5bdb34568

Author: Michael Chan <[email protected]>
Date:   Thu Sep 6 12:04:29 2007 +0100

[TG3]: Workaround MSI bug on 5714/5780.

A hardware bug was revealed after a recent PCI MSI patch was made to always disable legacy INTX when enabling MSI. The 5714/5780 chips will not generate MSI when INTX is disabled, causing MSI failure messages to be reported, and another patch was made to workaround the problem by disabling MSI on ServerWorks HT1000 bridge chips commonly found with the 5714.

We workaround this chip bug by enabling INTX after we enable MSI and after we resume from suspend.

Update version to 3.81.

This problem was discovered by David Miller.

hmm, I wonder if it would be a good idea to put together a git hook script that looks for things that look like git commit ID's, and if they aren't valid commit ID's that appear in the repository, the maintainer gets a warning when they do a "git am" or otherwise suck in a patch that was sent via e-mail...

- Ted

Well, the thing is, sometimes it makes sense.

In the stable tree, for example, the commits often point to the particular commit in the *development* tree that the particular stable commit was cherry-picked from. And that all makes perfect sense - but such a commit will not even exist in that tree (very much by definition: the whole point of a stable tree is to *not* have all the development commits in that tree, so just individual commits get moved over).

So it does make sense to point to commits in totally independent trees at times.

Yes, it could be an option, and yes, you could probably enable/disable it on a per-repository basis (ie the above kind of thing tens to make sense for certain repositories but not others). But it's definitely not something that necessarily always makes sense to do, so it's likely not a good default (and if it's not a default, then mistakes will continue to happen).

Linus

I'll defer to David on this one, as he sent it to me.

David?

greg k-h

Btw, I'd also like to take the opportunitity to ask people to always include the explanation of the commit when you give a commit ID.

"git revert" does that for the commit it reverts, but the (bogus) commit ID that was apparently added by David (c129d962a) didn't get the explanation for it, so now it's totally impossible to match it up with anything at all..

So let me one more say:

- don't just say

.. was cured by commit xyz

- but add the one-liner description of the commit, something along the lines of

.. was cured by commit xyz ("PCI: Add quirk for devices which disable MSI when INTX_DISABLE is set.")

so that the commit messages are readable even outside git.

Linus

From: Linus Torvalds <[email protected]>
Date: Mon, 5 Nov 2007 14:37:31 -0800 (PST)

I agree, and I've been trying to get into the habit of doing this even though I didn't this time.

From: Greg KH <[email protected]>
Date: Mon, 5 Nov 2007 14:17:52 -0800

At the time I originally wrote that patch, that was the commit ID in my net-2.6 tree, but I rebased net-2.6 before final submittion of the tg3 patch and I didn't update the commit log for this patch to match.

Sorry.