From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Return-Path: Date: Mon, 9 Oct 2017 08:54:51 +0200 From: Thomas Petazzoni To: Bjorn Helgaas Cc: Bjorn Helgaas , linux-pci@vger.kernel.org, Jason Cooper , Andrew Lunn , Sebastian Hesselbarth , Gregory Clement , Nadav Haklai , Hanna Hawa , Yehuda Yitschak , linux-arm-kernel@lists.infradead.org, Antoine Tenart , =?UTF-8?B?TWlxdcOobA==?= Raynal , Victor Gu Subject: Re: [PATCH v2 5/7] PCI: aardvark: disable LOS state by default Message-ID: <20171009085451.5a6cfbcc@windsurf.lan> In-Reply-To: <20171005174607.GS25517@bhelgaas-glaptop.roam.corp.google.com> References: <20170928125838.11887-1-thomas.petazzoni@free-electrons.com> <20170928125838.11887-6-thomas.petazzoni@free-electrons.com> <20171005174607.GS25517@bhelgaas-glaptop.roam.corp.google.com> MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII List-ID: Hello, On Thu, 5 Oct 2017 12:46:07 -0500, Bjorn Helgaas wrote: > On Thu, Sep 28, 2017 at 02:58:36PM +0200, Thomas Petazzoni wrote: > > From: Victor Gu > > > > Some PCIe devices do not support LOS, and will cause timeouts if the > > root complex forces the LOS state. This patch disables the LOS state > > by default. > > Per PCIe r3.1, sec 5.4.1.3, software should not enable L0s in either > direction unless both ends support L0s. > > I'm unclear on what the bug is here. Is the generic ASPM code > incorrectly enabling L0s when one end doesn't support it? That seems > unlikely, but if so, it should be fixed in the generic ASPM code. > > Are both ends advertising L0s support, but there's a hardware erratum > on the Aardvark end that keeps it from working correctly? If so, we > should say that explicitly and include a reference to a published > hardware erratum. > > Something else? I'll do some more research on this and get back to you. > > This is part of fixing bug > > https://bugzilla.kernel.org/show_bug.cgi?id=196339, this commit was > > reported as the user to be important to get a Intel 7260 mini-PCIe > > WiFi card working. > > The bugzilla link is a good start, but "reported by the user to be > important" is meaningless by itself. What we need here is the details > of what's broken and how this fixes it. Unfortunately the bugzilla > doesn't have those details. Yes, the issue is that the bug report doesn't have much details, and I don't have the specific PCIe card that was used by the bug reporter. Best regards, Thomas -- Thomas Petazzoni, CTO, Free Electrons Embedded Linux and Kernel engineering http://free-electrons.com