From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-0.8 required=3.0 tests=DKIMWL_WL_HIGH,DKIM_SIGNED, DKIM_VALID,HEADER_FROM_DIFFERENT_DOMAINS,MAILING_LIST_MULTI,SPF_HELO_NONE, SPF_PASS,URIBL_BLOCKED autolearn=no autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id A8F35C433DF for ; Tue, 7 Jul 2020 13:35:51 +0000 (UTC) Received: from merlin.infradead.org (merlin.infradead.org [205.233.59.134]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPS id 677AC20658 for ; Tue, 7 Jul 2020 13:35:51 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=pass (2048-bit key) header.d=lists.infradead.org header.i=@lists.infradead.org header.b="p6exhbx3"; dkim=fail reason="signature verification failed" (1024-bit key) header.d=chromium.org header.i=@chromium.org header.b="IJSp4ObV" DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 677AC20658 Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=chromium.org Authentication-Results: mail.kernel.org; spf=none smtp.mailfrom=linux-mediatek-bounces+linux-mediatek=archiver.kernel.org@lists.infradead.org DKIM-Signature: v=1; a=rsa-sha256; q=dns/txt; c=relaxed/relaxed; d=lists.infradead.org; s=merlin.20170209; h=Sender:Content-Transfer-Encoding: Content-Type:Cc:List-Subscribe:List-Help:List-Post:List-Archive: List-Unsubscribe:List-Id:To:Subject:Message-ID:Date:From:In-Reply-To: References:MIME-Version:Reply-To:Content-ID:Content-Description:Resent-Date: Resent-From:Resent-Sender:Resent-To:Resent-Cc:Resent-Message-ID:List-Owner; bh=pDRgfNSAwH1hxmRVIZXzHu+Au05jKPDLVVrWDc+WQIs=; b=p6exhbx3kYgOszle9OlB60F/A O1Rd6nprH7fXFJQQXMxw4y785J6ksGv59dai/9kjrIC4D5Vx8vfjNz/+vrYqS0KtJcz2+RH1Mx879 zvbZhEO1TzQdlxEeW+DZYiLhfuPQ2L+VoV41R8dul3U7s1NP3dl6e4UUaQzjH8mL1RaHkLKNQrJ0+ bhd6CuEmraFXeWOUt/DIyqgBL5kNOhSdVnN2bZ1uVtXlz0cwzMp4TVidKj2EpwQ3/cXfeun2XRf1a bjc5ug+Wd6+o727nbbo8eHdnaA1FD+bWe1oPpVIPjd7ZJ52trV3XO3YCqyGbJNJbsssX6z7X2nsmm gAtWxlLUA==; Received: from localhost ([::1] helo=merlin.infradead.org) by merlin.infradead.org with esmtp (Exim 4.92.3 #3 (Red Hat Linux)) id 1jsnl5-0003xU-Pd; Tue, 07 Jul 2020 13:35:39 +0000 Received: from mail-ej1-x642.google.com ([2a00:1450:4864:20::642]) by merlin.infradead.org with esmtps (Exim 4.92.3 #3 (Red Hat Linux)) id 1jsnl0-0003vU-7O for linux-mediatek@lists.infradead.org; Tue, 07 Jul 2020 13:35:35 +0000 Received: by mail-ej1-x642.google.com with SMTP id lx13so27896127ejb.4 for ; Tue, 07 Jul 2020 06:35:33 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=chromium.org; s=google; h=mime-version:references:in-reply-to:from:date:message-id:subject:to :cc; bh=6wQjqeHrYmelh0ahvaUo5qZ3Zhlm4Z5A2lb/5qO36ts=; b=IJSp4ObVCBA2+B4/0/8ipyQif2GwwqL1UopA4CTGsap4ni2MEvKpmZ2+eDoSabIp52 +gCy0fyqjd6exRFzAewcmhArVGV20XmF9LVOg5rn6LR7aW6rrTbVsxiKvaowSgO4Q9g6 Ki7QF70zXNWcDEWPfw52VDYf3sKoDSqHzX1Y0= X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:mime-version:references:in-reply-to:from:date :message-id:subject:to:cc; bh=6wQjqeHrYmelh0ahvaUo5qZ3Zhlm4Z5A2lb/5qO36ts=; b=IiqkV5UrrOurkRbVrl0h+CgyorfEO2Udj+SThYr2OjCLni2JCYVJFkG4S8DVfZSDVP V0tfJStH5waVBNa7QEUZ89juvh14+ekrVc/mBHSX6NJ40Y9nuI+eAUgntobFR3o82m42 NXgMSwARfvvbFGQIzFKgYUp/6+gJ7GDJ4Nwl4gvm0QM/XKDT1wx0WJy/9EDjkvgcKanW ZF8EJ/aIZgSGPiRMkrQsdGtbS1LwePVJMhdqDaKzUFaSlfIpZ9yQWzISCIPp040DiGNc pf4Xfps8QNgGwB15lAE3+4e2HHFJbMomxolQBvP7jDFdY1S2JEvyLnXQCR7AzCwLxXgi Q9fg== X-Gm-Message-State: AOAM5335VPdq/SBynCn1cAZNi9p2FRhX1NTfsb0/UaQKnteA8lREer33 q6iLl5TgYGU+3uZO97ohb+oylnhSRkDrWQ== X-Google-Smtp-Source: ABdhPJyHA2dSUUFzdmC7iiCzisqPZFdGK2tSUtDrFdOuZ0AEtvahTZT67oTPHXRnkBGGx1stddI17g== X-Received: by 2002:a17:907:72cf:: with SMTP id du15mr46013475ejc.151.1594128932109; Tue, 07 Jul 2020 06:35:32 -0700 (PDT) Received: from mail-wr1-f41.google.com (mail-wr1-f41.google.com. [209.85.221.41]) by smtp.gmail.com with ESMTPSA id dm1sm404966ejc.99.2020.07.07.06.35.30 for (version=TLS1_3 cipher=TLS_AES_128_GCM_SHA256 bits=128/128); Tue, 07 Jul 2020 06:35:31 -0700 (PDT) Received: by mail-wr1-f41.google.com with SMTP id r12so45097244wrj.13 for ; Tue, 07 Jul 2020 06:35:30 -0700 (PDT) X-Received: by 2002:adf:dfcd:: with SMTP id q13mr53974562wrn.295.1594128930262; Tue, 07 Jul 2020 06:35:30 -0700 (PDT) MIME-Version: 1.0 References: <20200604090553.10861-1-xia.jiang@mediatek.com> <20200604090553.10861-20-xia.jiang@mediatek.com> <20200611184640.GC8694@chromium.org> <1593485781.20112.43.camel@mhfsdcap03> <20200630165301.GA1212092@chromium.org> <1594104314.4473.24.camel@mhfsdcap03> In-Reply-To: <1594104314.4473.24.camel@mhfsdcap03> From: Tomasz Figa Date: Tue, 7 Jul 2020 15:35:18 +0200 X-Gmail-Original-Message-ID: Message-ID: Subject: Re: [PATCH RESEND v9 18/18] media: platform: Add jpeg enc feature To: Xia Jiang X-CRM114-Version: 20100106-BlameMichelson ( TRE 0.8.0 (BSD) ) MR-646709E3 X-CRM114-CacheID: sfid-20200707_093534_310009_92FD8C9D X-CRM114-Status: GOOD ( 43.57 ) X-BeenThere: linux-mediatek@lists.infradead.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: Nicolas Boichat , linux-devicetree , mojahsu@chromium.org, srv_heupstream , Rick Chang , Sergey Senozhatsky , Linux Kernel Mailing List , =?UTF-8?B?TWFvZ3VhbmcgTWVuZyAo5a2f5q+b5bm/KQ==?= , Mauro Carvalho Chehab , Sj Huang , Rob Herring , Matthias Brugger , Hans Verkuil , "moderated list:ARM/Mediatek SoC support" , Marek Szyprowski , "list@263.net:IOMMU DRIVERS , Joerg Roedel , " , Linux Media Mailing List Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Sender: "Linux-mediatek" Errors-To: linux-mediatek-bounces+linux-mediatek=archiver.kernel.org@lists.infradead.org On Tue, Jul 7, 2020 at 8:47 AM Xia Jiang wrote: > > On Tue, 2020-06-30 at 16:53 +0000, Tomasz Figa wrote: > > Hi Xia, > > > > On Tue, Jun 30, 2020 at 10:56:21AM +0800, Xia Jiang wrote: > > > On Thu, 2020-06-11 at 18:46 +0000, Tomasz Figa wrote: > > > > Hi Xia, > > > > > > > > On Thu, Jun 04, 2020 at 05:05:53PM +0800, Xia Jiang wrote: > > [snip] > > > > > +static void mtk_jpeg_enc_device_run(void *priv) > > > > > +{ > > > > > + struct mtk_jpeg_ctx *ctx = priv; > > > > > + struct mtk_jpeg_dev *jpeg = ctx->jpeg; > > > > > + struct vb2_v4l2_buffer *src_buf, *dst_buf; > > > > > + enum vb2_buffer_state buf_state = VB2_BUF_STATE_ERROR; > > > > > + unsigned long flags; > > > > > + struct mtk_jpeg_src_buf *jpeg_src_buf; > > > > > + struct mtk_jpeg_enc_bs enc_bs; > > > > > + int ret; > > > > > + > > > > > + src_buf = v4l2_m2m_next_src_buf(ctx->fh.m2m_ctx); > > > > > + dst_buf = v4l2_m2m_next_dst_buf(ctx->fh.m2m_ctx); > > > > > + jpeg_src_buf = mtk_jpeg_vb2_to_srcbuf(&src_buf->vb2_buf); > > > > > + > > > > > + ret = pm_runtime_get_sync(jpeg->dev); > > > > > + if (ret < 0) > > > > > + goto enc_end; > > > > > + > > > > > + spin_lock_irqsave(&jpeg->hw_lock, flags); > > > > > + > > > > > + /* > > > > > + * Resetting the hardware every frame is to ensure that all the > > > > > + * registers are cleared. This is a hardware requirement. > > > > > + */ > > > > > + mtk_jpeg_enc_reset(jpeg->reg_base); > > > > > + > > > > > + mtk_jpeg_set_enc_dst(ctx, jpeg->reg_base, &dst_buf->vb2_buf, &enc_bs); > > > > > + mtk_jpeg_set_enc_src(ctx, jpeg->reg_base, &src_buf->vb2_buf); > > > > > + mtk_jpeg_enc_set_config(jpeg->reg_base, ctx->out_q.fmt->hw_format, > > > > > + ctx->enable_exif, ctx->enc_quality, > > > > > + ctx->restart_interval); > > > > > + mtk_jpeg_enc_start(jpeg->reg_base); > > > > > > > > Could we just move the above 5 functions into one function inside > > > > mtk_jpeg_enc_hw.c that takes mtk_jpeg_dev pointer as its argument, let's > > > > say mtk_jpeg_enc_hw_run() and simply program all the data to the registers > > > > directly, without the extra level of abstractions? > > > I can move the 5 functions into one function(mtk_jpeg_enc_hw_run()), but > > > this function will be very long, because it contains computation code > > > such as setting dst addr, blk_num, quality. > > > In v4, you have adviced the following architecture: > > > How about the following model, as used by many other drivers: > > > > > > mtk_jpeg_enc_set_src() > > > { > > > // Set any registers related to source format and buffer > > > } > > > > > > mtk_jpeg_enc_set_dst() > > > { > > > // Set any registers related to destination format and buffer > > > } > > > > > > mtk_jpeg_enc_set_params() > > > { > > > // Set any registers related to additional encoding parameters > > > } > > > > > > mtk_jpeg_enc_device_run(enc, ctx) > > > { > > > mtk_jpeg_enc_set_src(enc, src_buf, src_fmt); > > > mtk_jpeg_enc_set_dst(enc, dst_buf, dst_fmt); > > > mtk_jpeg_enc_set_params(enc, ctx); > > > // Trigger the hardware run > > > } > > > I think that this architecture is more clear(mtk_jpeg_enc_set_config is > > > equivalent to mtk_jpeg_enc_set_params). > > > Should I keep the original architecture or move 5 functions into > > > mtk_jpeg_enc_hw_run? > > > > Sounds good to me. > > > > My biggest issue with the code that it ends up introducing one more > > level of abstraction, but with the approach you suggested, the arguments > > just accept standard structs, which avoids that problem. > > > > [snip] > > > > > + > > > > > + ctx->fh.ctrl_handler = &ctx->ctrl_hdl; > > > > > + ctx->colorspace = V4L2_COLORSPACE_JPEG, > > > > > + ctx->ycbcr_enc = V4L2_YCBCR_ENC_DEFAULT; > > > > > + ctx->quantization = V4L2_QUANTIZATION_DEFAULT; > > > > > + ctx->xfer_func = V4L2_XFER_FUNC_DEFAULT; > > > > > > > > Since we already have a v4l2_pix_format_mplane struct which has fields for > > > > the above 4 values, could we just store them there? > > > > > > > > Also, I don't see this driver handling the colorspaces in any way, but it > > > > seems to allow changing them from the userspace. This is incorrect, because > > > > the userspace has no way to know that the colorspace is not handled. > > > > Instead, the try_fmt implementation should always override the > > > > userspace-provided colorspace configuration with the ones that the driver > > > > assumes. > > > > > > > > > + pix_mp->width = MTK_JPEG_MIN_WIDTH; > > > > > + pix_mp->height = MTK_JPEG_MIN_HEIGHT; > > > > > + > > > > > + q->fmt = mtk_jpeg_find_format(V4L2_PIX_FMT_YUYV, > > > > > + MTK_JPEG_FMT_FLAG_ENC_OUTPUT); > > > > > + vidioc_try_fmt(container_of(pix_mp, struct v4l2_format, > > > > > + fmt.pix_mp), q->fmt); > > > > > + q->w = pix_mp->width; > > > > > + q->h = pix_mp->height; > > > > > + q->crop_rect.width = pix_mp->width; > > > > > + q->crop_rect.height = pix_mp->height; > > > > > + q->sizeimage[0] = pix_mp->plane_fmt[0].sizeimage; > > > > > + q->bytesperline[0] = pix_mp->plane_fmt[0].bytesperline; > > > > > > > > Actually, do we need this custom mtk_jpeg_q_data struct? Why couldn't we > > > > just keep the same values inside the standard v4l2_pix_format_mplane > > > > struct? > > > I think that we need mtk_jpeg_q_data struct.If we delete it, how can we > > > know these values(w, h, sizeimage, bytesperline, mtk_jpeg_fmt) belong to > > > output or capture(output and capture's sizeimages are different, width > > > and height are differnt too for jpeg dec )?We have > > > s_fmt_vid_out_mplane/cap_mplane function to set these values. > > > > > > But we can use standard v4l2_pix_format_mplane struct replacing the w, h > > > bytesperline, sizeimage in mtk_jpeg_q_data struct like this: > > > struct mtk_jpeg_q_data{ > > > struct mtk_jpeg_fmt *fmt; > > > struct v4l2_pix_format_mplane pix_mp; > > > struct v4l2_rect enc_crop_rect > > > } > > > Then delete ctx->colorspace ctx->ycbcr_enc ctx->quantization > > > ctx->xfer_func, becuase v4l2_pix_format_mplane in q_data has contained > > > them and assign them for out_q and cap_q separately. > > > > > > WDYT? > > > > Sounds good to me. I was considering just making it like > > > > struct mtk_jpeg_ctx { > > struct mtk_jpeg_fmt *src_fmt; > > struct v4l2_pix_format_mplane src_pix_mp; > > struct v4l2_rect src_crop; > > > > struct mtk_jpeg_fmt *dst_fmt; > > struct v4l2_pix_format_mplane dst_pix_mp; > > struct v4l2_rect dst_crop; > > }; > > > > but I like your suggestion as well, as long as custom data structures > > are not used to store standard information. > > Dear Tomasz, > > I used the structure like below: > struct mtk_jpeg_q_data{ > struct mtk_jpeg_fmt *fmt; > struct v4l2_pix_format_mplane pix_mp; > struct v4l2_rect enc_crop_rect > } > Then delete ctx->colorspace ctx->ycbcr_enc ctx->quantization > ctx->xfer_func, becuase v4l2_pix_format_mplane in q_data has contained > them and assign them for out_q and cap_q separately. > > Then the v4l2_compliance test will fail, the fail log as below: > Format ioctls: > test VIDIOC_ENUM_FMT/FRAMESIZES/FRAMEINTERVALS: OK > test VIDIOC_G/S_PARM: OK (Not Supported) > test VIDIOC_G_FBUF: OK (Not Supported) > test VIDIOC_G_FMT: OK > test VIDIOC_TRY_FMT: OK > fail: v4l2-test-formats.cpp(836): > fmt_cap.g_colorspace() != col > test VIDIOC_S_FMT: FAIL > test VIDIOC_G_SLICED_VBI_CAP: OK (Not Supported) > test Cropping: OK > test Composing: OK (Not Supported) > test Scaling: OK (Not Supported) > > The source code of v4l2-test-formats.cpp as below: > > static int testM2MFormats(struct node *node) > { > cv4l_fmt fmt_out; > cv4l_fmt fmt; > cv4l_fmt fmt_cap; > __u32 cap_type = node->g_type(); > __u32 out_type = v4l_type_invert(cap_type); > __u32 col, ycbcr_enc, quant, xfer_func; > > fail_on_test(node->g_fmt(fmt_out, out_type)); > node->g_fmt(fmt_cap, cap_type); > fail_on_test(fmt_cap.g_colorspace() != fmt_out.g_colorspace()); > fail_on_test(fmt_cap.g_ycbcr_enc() != fmt_out.g_ycbcr_enc()); > fail_on_test(fmt_cap.g_quantization() != fmt_out.g_quantization()); > fail_on_test(fmt_cap.g_xfer_func() != fmt_out.g_xfer_func()); > col = fmt_out.g_colorspace() == V4L2_COLORSPACE_SMPTE170M ? > V4L2_COLORSPACE_REC709 : V4L2_COLORSPACE_SMPTE170M; > ycbcr_enc = fmt_out.g_ycbcr_enc() == V4L2_YCBCR_ENC_601 ? > V4L2_YCBCR_ENC_709 : V4L2_YCBCR_ENC_601; > quant = fmt_out.g_quantization() == V4L2_QUANTIZATION_LIM_RANGE ? > V4L2_QUANTIZATION_FULL_RANGE : V4L2_QUANTIZATION_LIM_RANGE; > xfer_func = fmt_out.g_xfer_func() == V4L2_XFER_FUNC_SRGB ? > V4L2_XFER_FUNC_709 : V4L2_XFER_FUNC_SRGB; > fmt_out.s_colorspace(col); > fmt_out.s_xfer_func(xfer_func); > fmt_out.s_ycbcr_enc(ycbcr_enc); > fmt_out.s_quantization(quant); > node->s_fmt(fmt_out); > fail_on_test(fmt_out.g_colorspace() != col); > fail_on_test(fmt_out.g_xfer_func() != xfer_func); > fail_on_test(fmt_out.g_ycbcr_enc() != ycbcr_enc); > fail_on_test(fmt_out.g_quantization() != quant); > node->g_fmt(fmt_cap); > fail_on_test(fmt_cap.g_colorspace() != col); // line 836 > fail_on_test(fmt_cap.g_xfer_func() != xfer_func); > fail_on_test(fmt_cap.g_ycbcr_enc() != ycbcr_enc); > fail_on_test(fmt_cap.g_quantization() != quant); > } > > It needs that cap's colorspace equals out's colorspace when userspace > just set out's colorspace and then get cap's colorspace. However, cap's > colorspace which getted from driver equals V4L2_COLORSPACE_JPEG, because > the code in g_fmt() like this: > pix_mp->colorspace = q_data->pix_mp.colorspace; > pix_mp->ycbcr_enc = q_data->pix_mp.ycbcr_enc; > pix_mp->xfer_func = q_data->pix_mp.xfer_func; > pix_mp->quantization = q_data->pix_mp.quantization; > > How should I handle this case? Should I store them(colorspace, > sfer_func, ycbcr_enc, quatization) in ctx as the orinal desin? Then I > can get them from g_fmt() like this: > pix_mp->colorspace = ctx->colorspace; > pix_mp->ycbcr_enc = ctx->ycbcr_enc; > pix_mp->xfer_func = ctx->xfer_func; > pix_mp->quantization = ctx->quantization; Why would there be any other colorspace accepted? I suppose that the hardware only supports the JPEG color space, so it shouldn't accept any other colorspace in TRY_FMT (and thus S_FMT) anyway. Still, for correctness, I would suggest propagating the colorspace (and related) information from OUTPUT format to CAPTURE format in S_FMT(OUTPUT). Best regards, Tomasz > > Best Regards, > Xia Jiang > > [snip] > > > > > @@ -1042,8 +1619,12 @@ static int mtk_jpeg_probe(struct platform_device *pdev) > > > > > return jpeg_irq; > > > > > } > > > > > > > > > > - ret = devm_request_irq(&pdev->dev, jpeg_irq, mtk_jpeg_dec_irq, 0, > > > > > - pdev->name, jpeg); > > > > > + if (jpeg->variant->is_encoder) > > > > > + ret = devm_request_irq(&pdev->dev, jpeg_irq, mtk_jpeg_enc_irq, > > > > > + 0, pdev->name, jpeg); > > > > > + else > > > > > + ret = devm_request_irq(&pdev->dev, jpeg_irq, mtk_jpeg_dec_irq, > > > > > + 0, pdev->name, jpeg); > > > > > > > > Rather than having "is_encoder" in the variant struct, would it make more > > > > sense to have "irq_handler" instead? That would avoid the explicit if. > > > Do you mean to delete "is_encoder"? It is used 8 times in the > > > driver.Should I move them all to the match data? > > > > Yes. It would make the code linear and the varability between the > > decoder and encoder would be self-contained in the variant struct. > > > > Best regards, > > Tomasz > _______________________________________________ Linux-mediatek mailing list Linux-mediatek@lists.infradead.org http://lists.infradead.org/mailman/listinfo/linux-mediatek