Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757961AbYJIEql (ORCPT ); Thu, 9 Oct 2008 00:46:41 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1751398AbYJIEqb (ORCPT ); Thu, 9 Oct 2008 00:46:31 -0400 Received: from g4t0015.houston.hp.com ([15.201.24.18]:29459 "EHLO g4t0015.houston.hp.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750852AbYJIEqa (ORCPT ); Thu, 9 Oct 2008 00:46:30 -0400 From: Alex Chiang Subject: [PATCH v5 00/16] PCI: let the core manage slot names To: jbarnes@virtuousgeek.org, kristen.c.accardi@intel.com, matthew@wil.cx, kaneshige.kenji@jp.fujitsu.com, linux-pci@vger.kernel.org, linux-kernel@vger.kernel.org Date: Wed, 08 Oct 2008 22:46:29 -0600 Message-ID: <20081009043140.8678.44164.stgit@bob.kio> User-Agent: StGIT/0.14.3.215.gff3d MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit X-Brightmail-Tracker: AAAAAQAAAAI= X-Whitelist: TRUE Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 4984 Lines: 122 Hi all, This is v5 of the series that implements a series of changes that allows the PCI core to manage slot names, rather than individual hotplug drivers. Thanks again to Kenji-san for continuing to point out that my previous attempts were racy and had horrible, complicated code. :) And thanks to Willy for reviewing my poor attempt at introducing locking. I've decided to do them both a favor and take a new approach that does not change any locking semantics in any of the existing interfaces today (pci_hp_register, pci_hp_deregister, pci_create_slot). Keep in mind that the problems we're trying to solve are: - possible duplicate slot names from firmware - allow hotplug drivers to override detection driver slot names The race that Kenji-san points out occurs during slot creation, where both a detection driver and hotplug driver try to create a slot at the same time, with different names. The solution is to combine slot creation _and_ slot rename into one atomic operation via an updated pci_create_slot() API. pci_create_slot is already serialized, so that's a good start. We add a 'rename' parameter to pci_create_slot. The hotplug drivers set it to 1. The detection drivers set it to 0. Now, when trying to create a new slot: - check to see if it's already been created - if yes, and we want a rename, then go ahead and rename it - if yes, and we don't care about the name, then just return - if no, then create the slot normally This logic is fully described in the changelog in patch 04/16. I believe it is a much simpler approach than the earlier crap I was producing. I tested this series with the pci_slot driver and fakephp dup_slots=1 with multiple loads and unloads in different order, and it all seemed to work. I _hope_ this is the last round. ;) Thanks! /ac v4 -> v5: - add 'rename' param to pci_create_slot - make use of 'rename' param in pci_create_slot - remove v4 serialization - remove crap false name collsion code from v2 and v3 - rpaphp uses kstrdup (Thanks to Pekka Enberg for suggestion) v3 -> v4: - Do not access hotplug_slot_name() before name initialization - Serialize pci_hp_register/deregister v2 -> v3: - incorporate Willy's code review comments - fix possible memory leak, pointed out by Rolf Eike Beer - make false name collision detection work for empty slots - add 'dup_slots' module_param to fakephp to help debug all this ;) v1 -> v2: - fix possible false name collisions --- Alex Chiang (16): PCI Hotplug: fakephp: add duplicate slot name debugging PCI: Hotplug core: remove 'name' PCI: shcphp: remove 'name' parameter PCI: SGI Hotplug: stop managing bss_hotplug_slot->name PCI: rpaphp: kmalloc/kfree slot->name directly PCI: pciehp: remove 'name' parameter PCI: ibmphp: stop managing hotplug_slot->name PCI: fakephp: remove 'name' parameter PCI: cpqphp: stop managing hotplug_slot->name PCI: cpci_hotplug: stop managing hotplug_slot->name PCI: acpiphp: remove 'name' parameter PCI, PCI Hotplug: introduce slot_name helpers PCI: prevent duplicate slot names PCI: update pci_create_slot() to take a 'rename' param PCI: rename pci_update_slot_number to pci_renumber_slot PCI Hotplug core: add 'name' param pci_hp_register interface drivers/acpi/pci_slot.c | 2 drivers/pci/hotplug/acpiphp.h | 9 +- drivers/pci/hotplug/acpiphp_core.c | 32 ++++--- drivers/pci/hotplug/cpci_hotplug.h | 6 + drivers/pci/hotplug/cpci_hotplug_core.c | 75 ++++++---------- drivers/pci/hotplug/cpci_hotplug_pci.c | 4 - drivers/pci/hotplug/cpqphp.h | 13 +-- drivers/pci/hotplug/cpqphp_core.c | 43 ++++----- drivers/pci/hotplug/fakephp.c | 26 ++++- drivers/pci/hotplug/ibmphp.h | 5 - drivers/pci/hotplug/ibmphp_ebda.c | 19 +--- drivers/pci/hotplug/pci_hotplug_core.c | 30 ++---- drivers/pci/hotplug/pciehp.h | 9 +- drivers/pci/hotplug/pciehp_core.c | 49 ++++------ drivers/pci/hotplug/pciehp_ctrl.c | 53 ++++++----- drivers/pci/hotplug/pciehp_hpc.c | 1 drivers/pci/hotplug/rpaphp_slot.c | 10 +- drivers/pci/hotplug/sgi_hotplug.c | 18 +--- drivers/pci/hotplug/shpchp.h | 9 +- drivers/pci/hotplug/shpchp_core.c | 52 ++++------- drivers/pci/hotplug/shpchp_ctrl.c | 48 +++++----- drivers/pci/slot.c | 150 ++++++++++++++++++++++++------- include/linux/pci.h | 9 +- include/linux/pci_hotplug.h | 11 +- 24 files changed, 359 insertions(+), 324 deletions(-) -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/