Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754698AbdCMTEr (ORCPT ); Mon, 13 Mar 2017 15:04:47 -0400 Received: from mga04.intel.com ([192.55.52.120]:41885 "EHLO mga04.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751964AbdCMTEb (ORCPT ); Mon, 13 Mar 2017 15:04:31 -0400 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.36,160,1486454400"; d="scan'208";a="76133018" Subject: Re: [RFC 2/2] fpga manager: Add fpga_mgr_firmware_stream API To: Alan Tull References: <1489105090-4996-1-git-send-email-yi1.li@linux.intel.com> <1489105090-4996-3-git-send-email-yi1.li@linux.intel.com> Cc: ming.lei@canonical.com, mcgrof@kernel.org, Greg Kroah-Hartman , atull , Moritz Fischer , linux-kernel , linux-fpga@vger.kernel.org From: "Li, Yi" Message-ID: <8a6cd982-cb8f-41df-f108-ca07b6fe48a5@linux.intel.com> Date: Mon, 13 Mar 2017 14:04:06 -0500 User-Agent: Mozilla/5.0 (Windows NT 6.3; WOW64; rv:45.0) Gecko/20100101 Thunderbird/45.8.0 MIME-Version: 1.0 In-Reply-To: Content-Type: text/plain; charset=utf-8; format=flowed Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 6614 Lines: 177 Thanks Alan for the comments. On 3/13/2017 1:00 PM, Alan Tull wrote: > On Thu, Mar 9, 2017 at 6:18 PM, wrote: > > Hi Yi, > > Thanks for your RFC. I believe this functionality is badly needed. > > I had a few comments about the chunk size and some nits about comments below... > >> From: Yi Li >> >> Add fpga_mgr_firmware_stream API, which can load and program firmware >> >> in trucks to FPGA instead of the whole file. >> >> Signed-off-by: Yi Li >> --- >> drivers/fpga/fpga-mgr.c | 77 +++++++++++++++++++++++++++++++++++++++++++ >> include/linux/fpga/fpga-mgr.h | 4 +++ >> 2 files changed, 81 insertions(+) >> >> diff --git a/drivers/fpga/fpga-mgr.c b/drivers/fpga/fpga-mgr.c >> index 3206a53..bb55b80 100644 >> --- a/drivers/fpga/fpga-mgr.c >> +++ b/drivers/fpga/fpga-mgr.c >> @@ -27,10 +27,15 @@ >> #include >> #include >> #include >> +#include >> >> static DEFINE_IDA(fpga_mgr_ida); >> static struct class *fpga_mgr_class; >> >> +static int streamsize = SZ_4K; >> +module_param(streamsize, int, 0664); >> +MODULE_PARM_DESC(streamsize, "buffer size for firmware streaming"); > I think we could fix the chunk size at PAGE_SIZE unless someone can > state a reason for needing something different. Below I have one > reason why we will sometimes need > PAGE_SIZE. > >> + >> /* >> * Call the low level driver's write_init function. This will do the >> * device-specific things to get the FPGA into the state where it is ready to >> @@ -309,6 +314,78 @@ int fpga_mgr_firmware_load(struct fpga_manager *mgr, >> } >> EXPORT_SYMBOL_GPL(fpga_mgr_firmware_load); >> >> +/** >> + * fpga_mgr_firmware_load - request firmware and load to fpga >> + * @mgr: fpga manager >> + * @info: fpga image specific information >> + * @image_name: name of image file on the firmware search path >> + * >> + * Request an FPGA image using the firmware class, then write out to the FPGA. >> + * Update the state before each step to provide info on what step failed if >> + * there is a failure. This code assumes the caller got the mgr pointer >> + * from of_fpga_mgr_get() or fpga_mgr_get() and checked that it is not an error >> + * code. >> + * >> + * Return: 0 on success, negative error code otherwise. >> + */ >> +int fpga_mgr_firmware_stream(struct fpga_manager *mgr, >> + struct fpga_image_info *info, >> + const char *image_name) > No need for both fpga_mgr_firmware_load and fpga_mgr_firmware_stream. > Just replace the old fpga_mgr_firmware_load with this function. Sure, will do. > >> +{ >> + struct device *dev = &mgr->dev; >> + const struct firmware *fw = NULL; >> + int ret; >> + size_t size = INT_MAX, offset = 0; >> + bool start_flag = 1; >> + >> + dev_info(dev, "writing %s to %s with buffer size %d\n", >> + image_name, mgr->name, streamsize); >> + >> + mgr->state = FPGA_MGR_STATE_FIRMWARE_REQ; >> + >> + while (size > 0) { >> + ret = stream_firmware(&fw, image_name, dev, offset, streamsize); >> + if (ret) { >> + dev_err(dev, "Error reading firmware %d\n", ret); >> + mgr->state = FPGA_MGR_STATE_FIRMWARE_REQ_ERR; >> + break; >> + } >> + >> + /* >> + * init. >> + */ > We don't really need this comment. > >> + if (start_flag) { >> + ret = fpga_mgr_write_init_buf(mgr, info, fw->data, >> + fw->size); > When the fpga manager is registered, one of the ops is > initial_header_size, which specifies how much buffer the write_init > needs. So this call could fail if initial_header_size < streamsize. > > I suggest that you make the streamsize equal to the smallest multiple > of PAGE_SIZE that is >= initial_header_size. Agree, we can fix the streamsize to something bigger than initial_header_size for the bitstream file. Another reason to have the variable streamsize is for the performance, although with Altera Cyclone V FPGA CvP I did not see much difference when increasing the buffer size, since the bottleneck is with programming through PCIe and the time spent with firmware file reading is negligible. > >> + start_flag = 0; >> + if (ret) >> + break; >> + } >> + >> + /* >> + * Write the FPGA image to the FPGA. >> + */ > Or this comment. > >> + mgr->state = FPGA_MGR_STATE_WRITE; >> + ret = mgr->mops->write(mgr, fw->data, fw->size); >> + if (ret) { >> + dev_err(dev, "Error while writing image data to FPGA\n"); >> + mgr->state = FPGA_MGR_STATE_WRITE_ERR; >> + break; >> + } >> + >> + offset += fw->size; >> + size -= fw->size; >> + if (fw->size < streamsize) >> + break; >> + } >> + >> + ret = fpga_mgr_write_complete(mgr, info); >> + >> + release_firmware(fw); >> + return ret; >> +} >> +EXPORT_SYMBOL_GPL(fpga_mgr_firmware_stream); >> + >> int fpga_mgr_load(struct fpga_manager *mgr, struct fpga_image_info *info) >> { >> if (info->firmware_name) >> diff --git a/include/linux/fpga/fpga-mgr.h b/include/linux/fpga/fpga-mgr.h >> index 0f5072c..a25362e 100644 >> --- a/include/linux/fpga/fpga-mgr.h >> +++ b/include/linux/fpga/fpga-mgr.h >> @@ -151,6 +151,10 @@ int fpga_mgr_firmware_load(struct fpga_manager *mgr, >> struct fpga_image_info *info, >> const char *image_name); >> >> +int fpga_mgr_firmware_stream(struct fpga_manager *mgr, >> + struct fpga_image_info *info, >> + const char *image_name); > Don't need both fpga_mgr_firmware_load and fpga_mgr_firmware_stream, > so no need to change the header. > >> + >> int fpga_mgr_load(struct fpga_manager *mgr, struct fpga_image_info *info); >> >> int fpga_mgr_lock(struct fpga_manager *mgr); >> -- >> 2.7.4 >> > -- > To unsubscribe from this list: send the line "unsubscribe linux-fpga" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html