From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Return-Path: From: "Jingoo Han" To: "'Felipe Balbi'" , "'Bjorn Helgaas'" Cc: "'Bjorn Helgaas'" , , References: <20180803065120.29322-1-felipe.balbi@linux.intel.com> <20180806182327.GA30691@bhelgaas-glaptop.roam.corp.google.com> <87lg9ia8q6.fsf@linux.intel.com> In-Reply-To: <87lg9ia8q6.fsf@linux.intel.com> Subject: Re: [PATCH] PCI: Fix bit definitions for LNKCAP2 register Date: Tue, 7 Aug 2018 14:18:23 -0400 Message-ID: <000001d42e7b$07990170$16cb0450$@gmail.com> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" List-ID: On Tuesday, August 7, 2018 4:21 AM, Felipe Balbi wrote: > > > Hi, > > Bjorn Helgaas writes: > > On Fri, Aug 03, 2018 at 09:51:20AM +0300, Felipe Balbi wrote: > >> Even thhough commit b891b4dc1eed claimed that original bit definitions > >> were wrong, that's not really the case. After verifying PCI > >> Specification Revisions 3.0, 3.1 and 4.0, Link Capabilites 2 > >> Register's bit definitions were always starting from Bit 0. > >> > >> This has been causing issues reporting correct link speeds on sysfs. > > > > Can you elaborate on this a bit? b891b4dc1eed still looks correct to > > me. I'm looking at PCIe r4.0, sec 7.5.3.18, where it shows: > > > > bit 0 RsvdP > > bits 7:1 Supported Link Speeds Vector > > I had missed this detail, actually. It was a misinterpretation of the > spec. Sorry for the noise. Hi Balbi, I can understand your situation. The detail of PCIe spec is very confusing. Actually, I was one of those who misunderstood. However, after reviewing the PCIe spec carefully, I found that the bit definition was wrong. At that time, I already checked that my patch works properly with Exynos SoCs. Thank you. Best regards, Jingoo Han > > -- > balbi