From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from li153-180.members.linode.com ([109.74.206.180]:36811 "EHLO mail.tekno-soft.it" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751925AbcCWLBz (ORCPT ); Wed, 23 Mar 2016 07:01:55 -0400 Subject: Re: iMX6q: Kernel panic when enabling interrupt for more than 2 cards behind a PCIe-to-PCI bridge To: Lucas Stach References: <56F1747A.7080602@tekno-soft.it> <1458728766.2171.39.camel@pengutronix.de> <56F2727F.8080205@tekno-soft.it> <1458730121.2171.44.camel@pengutronix.de> Cc: "linux-pci@vger.kernel.org" , Richard Zhu , Fabio Estevam From: Roberto Fichera Message-ID: <56F27792.9080600@tekno-soft.it> Date: Wed, 23 Mar 2016 12:01:38 +0100 MIME-Version: 1.0 In-Reply-To: <1458730121.2171.44.camel@pengutronix.de> Content-Type: text/plain; charset=utf-8 Sender: linux-pci-owner@vger.kernel.org List-ID: On 03/23/2016 11:48 AM, Lucas Stach wrote: > Am Mittwoch, den 23.03.2016, 11:39 +0100 schrieb Roberto Fichera: >> On 03/23/2016 11:26 AM, Lucas Stach wrote: >> >> Hi Lucas, >> >>> Am Mittwoch, den 23.03.2016, 07:13 -0300 schrieb Fabio Estevam: >>>> On Tue, Mar 22, 2016 at 1:36 PM, Roberto Fichera wrote: >>>>> Hi All, >>>>> >>>>> I'm getting a kernel panic with a kernel v4.4.x when enabling interrupt for 2 cards behind a PCIe-to-PCI bridge >>>>> not sharing the same IRQ. If they don't share the same IRQ most often I'm getting a message claiming the emmc >>>>> is stuck. I've disabled MSI from PCI config because the cards are not getting any interrupt delivered. >>> Is the interrupt-map on your PCIe-to-PCI bridge correct? If it isn't the >>> CPU may drown in unhandled IRQ storms. >> Can you please be more specific? Are you talking about INTA/B/C/D mapping? >> > Yes. If you are using legacy interrupts that are not shared between > devices I suppose you are using multiple IRQ pins. > > The interrupt-map on the i.MX6 PCI controller maps those legacy IRQs to > the internal ARM GIC IRQs, but it doesn't say anything about how your > bridge maps them. > > If it doesn't map them 1:1, you need to have an interrupt-map property > at the bridge to define how the mapping is done. Ok! Now I see! The bridge uses only INTA/B for the miniPCI slots > > Regards, > Lucas > > >