Messages in this thread | | | From | "Gupta, Nipun" <> | Subject | RE: [PATCH 01/19] bus/cdx: add the cdx bus driver | Date | Wed, 18 Jan 2023 10:56:27 +0000 |
| |
[AMD Official Use Only - General]
> -----Original Message----- > From: Greg KH <gregkh@linuxfoundation.org> > Sent: Tuesday, January 17, 2023 7:37 PM > To: Gupta, Nipun <Nipun.Gupta@amd.com> > Cc: robh+dt@kernel.org; krzysztof.kozlowski+dt@linaro.org; rafael@kernel.org; > eric.auger@redhat.com; alex.williamson@redhat.com; cohuck@redhat.com; > song.bao.hua@hisilicon.com; mchehab+huawei@kernel.org; maz@kernel.org; > f.fainelli@gmail.com; jeffrey.l.hugo@gmail.com; saravanak@google.com; > Michael.Srba@seznam.cz; mani@kernel.org; yishaih@nvidia.com; > jgg@ziepe.ca; jgg@nvidia.com; robin.murphy@arm.com; will@kernel.org; > joro@8bytes.org; masahiroy@kernel.org; ndesaulniers@google.com; linux-arm- > kernel@lists.infradead.org; linux-kbuild@vger.kernel.org; linux- > kernel@vger.kernel.org; devicetree@vger.kernel.org; okaya@kernel.org; > Anand, Harpreet <harpreet.anand@amd.com>; Agarwal, Nikhil > <nikhil.agarwal@amd.com>; Simek, Michal <michal.simek@amd.com>; git > (AMD-Xilinx) <git@amd.com> > Subject: Re: [PATCH 01/19] bus/cdx: add the cdx bus driver > > Caution: This message originated from an External Source. Use proper caution > when opening attachments, clicking links, or responding. > > > On Tue, Jan 17, 2023 at 07:11:33PM +0530, Nipun Gupta wrote: > > Introduce AMD CDX bus, which provides a mechanism for scanning > > and probing CDX devices. These devices are memory mapped on > > system bus for Application Processors(APUs). > > > > CDX devices can be changed dynamically in the Fabric and CDX > > bus interacts with CDX controller to rescan the bus and > > rediscover the devices. > > > > Signed-off-by: Nipun Gupta <nipun.gupta@amd.com> > > Signed-off-by: Tarak Reddy <tarak.reddy@amd.com> > > First off, very nice job, I didn't find any obvious issues with this > integration into the driver core. > > That being said, why do you want this in drivers/bus/? Why not > drivers/cdx/ ?
Thanks, Greg, for taking time out for review and providing your valuable feedback. We do not have strong affiliation to drivers/bus/cdx so will move it to drivers/cdx in the next spin.
> > One minor comment to make the code smaller: > > > +static int get_free_index(void) > > +{ > > + unsigned long id_map; > > + unsigned long mask; > > + int index = 0; > > + > > + mask = (1UL << MAX_CDX_CONTROLLERS) - 1; > > +retry: > > + id_map = cdx_controller_id_map[0]; > > + if ((id_map & mask) == mask) > > + return -ENOSPC; > > + > > + index = ffz(id_map); > > + if (index >= MAX_CDX_CONTROLLERS) > > + return -ENOSPC; > > + > > + if (test_and_set_bit(index, &cdx_controller_id_map[0])) > > + goto retry; > > + > > + return index; > > +} > > Why not just use the idr/ida structure instead? That will handle all of > that logic for you and get rid of your bit twiddling.
Agree. Using idr/ida seems more appropriate. Will update the code accordingly.
> > > +/** > > + * struct cdx_dev_params - CDX device parameters > > + * @cdx: CDX controller associated with the device > > + * @parent: Associated CDX controller > > + * @vendor: Vendor ID for CDX device > > + * @device: Device ID for CDX device > > + * @bus_num: Bus number for this CDX device > > + * @dev_num: Device number for this device > > + * @res: array of MMIO region entries > > + * @res_count: number of valid MMIO regions > > + * @req_id: Requestor ID associated with CDX device > > + */ > > +struct cdx_dev_params { > > + struct cdx_controller *cdx; > > + u16 vendor; > > + u16 device; > > Are these in little endian format in memory? Or native? Or something > else?
While reading from the hardware, these values are little-endian; and they are converted to CPU endianness by the controller code using le_to_cpu32() and then passed as CPU endian while registering the CDX device.
Thanks, Nipun
> > thanks, > > greg k-h
| |