Hi Zhiqiang, Às 4:48 AM de 7/7/2017, Z.q. Hou escreveu: > Hi Joao, > > > > Thanks a lot for your comments! > > > >> -----Original Message----- > >> From: Joao Pinto [mailto:Joao.Pinto@xxxxxxxxxxxx] > >> Sent: 2017年7月6日 17:44 > >> To: Z.q. Hou <zhiqiang.hou@xxxxxxx>; linux-pci@xxxxxxxxxxxxxxx; > >> Joao.Pinto@xxxxxxxxxxxx > >> Cc: bhelgaas@xxxxxxxxxx; jingoohan1@xxxxxxxxx > >> Subject: Re: [PATCH 1/3] PCI: designware: add accessors for write permission > >> of DBI read-only registers > >> > >> > >> Hi Zhiqiang, > >> > >> Às 7:33 AM de 7/6/2017, Zhiqiang Hou escreveu: > >>> From: Hou Zhiqiang <Zhiqiang.Hou@xxxxxxx> > >>> > >>> The read-only DBI registers can be written over the DBI when set the > >>> "Write to RO Registers Using DBI" (DBI_RO_WR_EN) field of the > >>> MISC_CONTROL_1_OFF register. > >> > >> I would suggest you to add a cover-letter next time to explain the global > >> picture of the patch-set. > > > > Thanks, I will. > > > >> > >> I understand your need for this patch, but I don't agree on the approach. > > > > In the DWC common code, there is a function write a DBI read-only register 'Device class code', and the first 2 patches is to fix it. > > The 3rd patch is to refactor the Layerscape PCIe driver's host_init function and reuse the new added accessors. > > > >> Sometimes the people in charge of the hardware design / configuration, forget > >> to specify the device class and that can be problematic for some drivers, and > >> so the typical workaround is to set it in the driver using a quirk for example. > >> > >> You can see some examples here: > >> https://urldefense.proofpoint.com/v2/url?u=https-3A__git.kernel.org_pub_scm_linux_kernel_git_torvalds_linux.git_tree_drivers_&d=DwIGaQ&c=DPL6_X_6JkXFx7AXWqB0tg&r=s2fO0hii0OGNOv9qQy_HRXy-xAJUD1NNoEcc3io_kx0&m=0DrBCvOc_J7LaMXBei1qCXxxfbLxWaVErKZ6Rkm6bUc&s=sOEmExQFqrCEmpAx9LjSeKRvkW1D-W82ckX5WGCgFWw&e= > >> pci/quirks.c > > > > I don't know the PCI quirks, do you mean remove the pci Device Class fix code from the DWC common code and add it to quirks? > In my opinion adding fixes to a common code is not a good approach. I would suggest the fix to go into the quirks file. @Bjorn: The quirks file is the best place for this type of fixes right? Thanks, Joao > > > Thanks, > > Zhiqiang >
