From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753524AbcFQQdV (ORCPT ); Fri, 17 Jun 2016 12:33:21 -0400 Received: from foss.arm.com ([217.140.101.70]:51763 "EHLO foss.arm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751768AbcFQQdT (ORCPT ); Fri, 17 Jun 2016 12:33:19 -0400 Date: Fri, 17 Jun 2016 17:33:27 +0100 From: Jean-Philippe Brucker To: Eric Auger Cc: eric.auger.pro@gmail.com, robin.murphy@arm.com, alex.williamson@redhat.com, will.deacon@arm.com, joro@8bytes.org, tglx@linutronix.de, jason@lakedaemon.net, marc.zyngier@arm.com, christoffer.dall@linaro.org, linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org, Bharat.Bhushan@freescale.com, pranav.sawargaonkar@gmail.com, p.fedin@samsung.com, iommu@lists.linux-foundation.org, julien.grall@arm.com, yehuday@marvell.com Subject: Re: [PATCH v10 7/9] irqchip/gicv3-its: register the MSI global doorbell Message-ID: <20160617163327.GA26522@e106794-lin.localdomain> References: <1465315288-5931-1-git-send-email-eric.auger@redhat.com> <1465315288-5931-8-git-send-email-eric.auger@redhat.com> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline In-Reply-To: <1465315288-5931-8-git-send-email-eric.auger@redhat.com> User-Agent: Mutt/1.6.1 (2016-04-27) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hi Eric, On Tue, Jun 07, 2016 at 04:01:26PM +0000, Eric Auger wrote: > This patch adds the registration of the MSI global doorbell in > gicv3-its driver. > > This will allow the msi layer to iommu_map this doorbell when > requested. > > Signed-off-by: Eric Auger > --- > drivers/irqchip/irq-gic-v3-its.c | 14 ++++++++++++++ > 1 file changed, 14 insertions(+) > > diff --git a/drivers/irqchip/irq-gic-v3-its.c b/drivers/irqchip/irq-gic-v3-its.c > index 5eb1f9e..ed9dfce 100644 > --- a/drivers/irqchip/irq-gic-v3-its.c > +++ b/drivers/irqchip/irq-gic-v3-its.c > @@ -29,6 +29,8 @@ > #include > #include > #include > +#include > +#include > > #include > #include > @@ -1607,6 +1609,7 @@ static int __init its_probe(struct device_node *node, > > if (of_property_read_bool(node, "msi-controller")) { > struct msi_domain_info *info; > + phys_addr_t translater; > > info = kzalloc(sizeof(*info), GFP_KERNEL); > if (!info) { > @@ -1614,10 +1617,21 @@ static int __init its_probe(struct device_node *node, > goto out_free_tables; > } > > + translater = its->phys_base + GITS_TRANSLATER; > + err = msi_doorbell_register_global(its, translater, > + sizeof(u32), > + IOMMU_WRITE | IOMMU_MMIO, true); This doesn't work :( First we have its_probe registering the global mapping with doorbell->chip_data = its, which is a pointer to an its_node structure. But when enabling MSIs for a device, its_irq_domain_alloc puts a pointer to an *its_device* into irq_data->chip_data. This seems to be a per-device structure, allocated by its_msi_prepare. The following call to msi_doorbell_lookup won't ever succeed, because it will compare its_node to its_device. I can't figure out how to fix it cleanly at the moment. Jean-Philippe > + if (err) { > + kfree(info); > + goto out_free_tables; > + } > + > + > inner_domain = irq_domain_add_tree(node, &its_domain_ops, its); > if (!inner_domain) { > err = -ENOMEM; > kfree(info); > + msi_doorbell_unregister(its); > goto out_free_tables; > } > > -- > 1.9.1 >