From patchwork Thu Dec 5 09:24:50 2019 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Neil Armstrong X-Patchwork-Id: 11274385 Return-Path: Received: from mail.kernel.org (pdx-korg-mail-1.web.codeaurora.org [172.30.200.123]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id 864EC139A for ; Thu, 5 Dec 2019 09:25:00 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 63EDE2464F for ; Thu, 5 Dec 2019 09:25:00 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=pass (2048-bit key) header.d=baylibre-com.20150623.gappssmtp.com header.i=@baylibre-com.20150623.gappssmtp.com header.b="I/lfkLke" Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1729133AbfLEJY7 (ORCPT ); Thu, 5 Dec 2019 04:24:59 -0500 Received: from mail-wr1-f66.google.com ([209.85.221.66]:37395 "EHLO mail-wr1-f66.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1729131AbfLEJY7 (ORCPT ); Thu, 5 Dec 2019 04:24:59 -0500 Received: by mail-wr1-f66.google.com with SMTP id w15so2593813wru.4 for ; Thu, 05 Dec 2019 01:24:58 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=baylibre-com.20150623.gappssmtp.com; s=20150623; h=from:to:cc:subject:date:message-id:in-reply-to:references :mime-version:content-transfer-encoding; bh=nq23CNv+1Mccfg53lZf4BbmhYzh0ats8L7b+6kh1lhM=; b=I/lfkLkec9xB4Lt64wD0JnmL1JPuqJcyEsa4h0hoXOUU/LWyELsD2B6FcUs8hnngLr s+iXCwOz//+2tsj4q3jUPAYlFnr0S97iUKEIVw+bh962MrKywBxIUjTHjF4VGEmEXxVu CblxxSfJpk6JGELdeXTC1lPCXGZpZRt1miq63dr0M1xsDi7IUpJfGpOe/EuMDXnbgnJM Qc58/qbQsGckCR1Kc1OD/KYg1fDFqt+cCX/Kijan7HudoaA1WfcMQMK+a/xiAv0p67PL 4/gFjYTNRyxXZCX+yStyq4ucTXOkp3V/Ye3YdM2rquaNhy/4TNzxz3RKEkPx3e6p5Va5 ok6w== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to :references:mime-version:content-transfer-encoding; bh=nq23CNv+1Mccfg53lZf4BbmhYzh0ats8L7b+6kh1lhM=; b=eycNJ3JZMcTX81cLHTphKC07ZDfG4j1IY0ZIFTbg9rY/4DDc59jd3NIIiLAlfxkIQm BrsuN1jOlUbjVLd10luxXF381OS0tNYwyq0Ze7gpGQyG8eDQruBuTAsabpXDWXhJ4Y9q rYJsy6fGdqVUDdzrgqIBOCDIyvAWdC+wmy7Nf7pUW9Y7CJ2geyMDSW7RxtXPeAfsra95 yK3DGDMJy8JMtDiai8YU0xHrf1uo5EIYuwpbFmEccxcAb4a/MSTvjSvmHnXzHfXVxFqP EYQPD7RKeW9q7JsdYmNYOT3XMYFe8ImMc9DuOYxFJMGzUGP0TJ5iaRuqM+HU3WbdhnON IfYA== X-Gm-Message-State: APjAAAXY0waTyaH7su8zPdwepOHYCVtp8iRfHF1+qqdflVtDRdz1Diey N6gOZmTTygl1CaXIuFAM1rNMgrKwYvwK5w== X-Google-Smtp-Source: APXvYqxM/twH277jtVYnEriu835IS5kknyb1j1etNDY9mX3nYl9pJof1Rjwetb+ekuL3yMja3982dQ== X-Received: by 2002:a5d:4752:: with SMTP id o18mr8664381wrs.330.1575537897807; Thu, 05 Dec 2019 01:24:57 -0800 (PST) Received: from bender.baylibre.local (wal59-h01-176-150-251-154.dsl.sta.abo.bbox.fr. [176.150.251.154]) by smtp.gmail.com with ESMTPSA id t5sm11642620wrr.35.2019.12.05.01.24.57 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Thu, 05 Dec 2019 01:24:57 -0800 (PST) From: Neil Armstrong To: mchehab@kernel.org, hans.verkuil@cisco.com Cc: Neil Armstrong , linux-media@vger.kernel.org, linux-amlogic@lists.infradead.org, linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org Subject: [PATCH 1/5] media: meson: vdec: align stride on 32 bytes Date: Thu, 5 Dec 2019 10:24:50 +0100 Message-Id: <20191205092454.26075-2-narmstrong@baylibre.com> X-Mailer: git-send-email 2.22.0 In-Reply-To: <20191205092454.26075-1-narmstrong@baylibre.com> References: <20191205092454.26075-1-narmstrong@baylibre.com> MIME-Version: 1.0 Sender: linux-media-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-media@vger.kernel.org The HEVC/VP9 aligns the plane stride on 32, so align the planes stride for all codecs to 32 to satisfy HEVC/VP9 decoding using the "HEVC" HW. This fixes VP9 decoding of streams with following (not limited) widths: - 264 -288 - 350 - 352 - 472 - 480 - 528 - 600 - 720 - 800 - 848 - 1440 Signed-off-by: Neil Armstrong --- drivers/staging/media/meson/vdec/vdec.c | 10 +++++----- drivers/staging/media/meson/vdec/vdec_helpers.c | 4 ++-- 2 files changed, 7 insertions(+), 7 deletions(-) diff --git a/drivers/staging/media/meson/vdec/vdec.c b/drivers/staging/media/meson/vdec/vdec.c index f03d9eba6e23..bd59d32b92ae 100644 --- a/drivers/staging/media/meson/vdec/vdec.c +++ b/drivers/staging/media/meson/vdec/vdec.c @@ -525,20 +525,20 @@ vdec_try_fmt_common(struct amvdec_session *sess, u32 size, memset(pfmt[1].reserved, 0, sizeof(pfmt[1].reserved)); if (pixmp->pixelformat == V4L2_PIX_FMT_NV12M) { pfmt[0].sizeimage = output_size; - pfmt[0].bytesperline = ALIGN(pixmp->width, 64); + pfmt[0].bytesperline = ALIGN(pixmp->width, 32); pfmt[1].sizeimage = output_size / 2; - pfmt[1].bytesperline = ALIGN(pixmp->width, 64); + pfmt[1].bytesperline = ALIGN(pixmp->width, 32); pixmp->num_planes = 2; } else if (pixmp->pixelformat == V4L2_PIX_FMT_YUV420M) { pfmt[0].sizeimage = output_size; - pfmt[0].bytesperline = ALIGN(pixmp->width, 64); + pfmt[0].bytesperline = ALIGN(pixmp->width, 32); pfmt[1].sizeimage = output_size / 4; - pfmt[1].bytesperline = ALIGN(pixmp->width, 64) / 2; + pfmt[1].bytesperline = ALIGN(pixmp->width, 32) / 2; pfmt[2].sizeimage = output_size / 2; - pfmt[2].bytesperline = ALIGN(pixmp->width, 64) / 2; + pfmt[2].bytesperline = ALIGN(pixmp->width, 32) / 2; pixmp->num_planes = 3; } } diff --git a/drivers/staging/media/meson/vdec/vdec_helpers.c b/drivers/staging/media/meson/vdec/vdec_helpers.c index ff4333074197..fc59d8801643 100644 --- a/drivers/staging/media/meson/vdec/vdec_helpers.c +++ b/drivers/staging/media/meson/vdec/vdec_helpers.c @@ -154,8 +154,8 @@ int amvdec_set_canvases(struct amvdec_session *sess, { struct v4l2_m2m_buffer *buf; u32 pixfmt = sess->pixfmt_cap; - u32 width = ALIGN(sess->width, 64); - u32 height = ALIGN(sess->height, 64); + u32 width = ALIGN(sess->width, 32); + u32 height = ALIGN(sess->height, 32); u32 reg_cur = reg_base[0]; u32 reg_num_cur = 0; u32 reg_base_cur = 0; From patchwork Thu Dec 5 09:24:51 2019 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Neil Armstrong X-Patchwork-Id: 11274391 Return-Path: Received: from mail.kernel.org (pdx-korg-mail-1.web.codeaurora.org [172.30.200.123]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id 4E5A214B7 for ; Thu, 5 Dec 2019 09:25:15 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 2C64624651 for ; Thu, 5 Dec 2019 09:25:15 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=pass (2048-bit key) header.d=baylibre-com.20150623.gappssmtp.com header.i=@baylibre-com.20150623.gappssmtp.com header.b="tP66xRoB" Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1729155AbfLEJZB (ORCPT ); Thu, 5 Dec 2019 04:25:01 -0500 Received: from mail-wm1-f68.google.com ([209.85.128.68]:36716 "EHLO mail-wm1-f68.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1729134AbfLEJZA (ORCPT ); Thu, 5 Dec 2019 04:25:00 -0500 Received: by mail-wm1-f68.google.com with SMTP id p17so2811941wma.1 for ; Thu, 05 Dec 2019 01:25:00 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=baylibre-com.20150623.gappssmtp.com; s=20150623; h=from:to:cc:subject:date:message-id:in-reply-to:references :mime-version:content-transfer-encoding; bh=Q8VsecRt+zfmEiQATRM3jixSlKuyuB/bDX0OonY2El4=; b=tP66xRoBmBGI/tMKDtGfs0UrFlQXywBHuH8Gm6UbBRUGdWZHbAepRHCsCIUWUbf61t PeAKi5xnpO20UTAZNoW985568KpI0/H9vnpy9txrTpa4yD+WA2523ZhlFwhxrJ6GIi78 TE1jSfYZX7p/Vd1idYyt5Xr3ylxedwKcX4iBVqRJ54vr7T+V/ObdMli0rpTZdyzCFIWP AxzUbnpSVIDxRjpO25I5UrjegOFTXC416HJb/9B0BuVl03vqgYMV4nJQ1Xq89hXJ2/yg JDdZZHauMJY/iEPM3Wgcla+xFymbyHoq2hzd5edjCp/f+qE1C6YSODgeJ1x/zfV51nsQ GlSg== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to :references:mime-version:content-transfer-encoding; bh=Q8VsecRt+zfmEiQATRM3jixSlKuyuB/bDX0OonY2El4=; b=tKIwdebwf1A3T88Nrn352UJ+EqbwTvDblWyTd2dYWSBTMpYU3ORT2doJx92+2KE4xZ SNPuuRJKz2sargsX5icfU7ekITcMLAtZFxRnGDr8XBctQ6IDWCM9AQc44nxXqMOsXPZp GaGh4zpfVLjAXlQWC/d2+iCFeTTqZUKJbBct3SrSm0FdxEC2brFr7FNDbi3eV6hwm3OA jWLm9RIvLPSJZqgMiW4u1A1w9M5oDY0JN4svYUGlb5L7weKepAcUhEDe07H4bXsr7+bu ml9Y2CIJDg3tDGVKymgdONIdhIvdTDYMe9LPdOzOefkgLJrGCHiI8PjYgzLvxbwADru0 1KGQ== X-Gm-Message-State: APjAAAV7b2Yz3r1RvzXiWKppUjxw3WxlqTOEl9oZrGf/iWQDc/JedxSi L67vrMfU8g5ZG6hCXElMcu7ikg== X-Google-Smtp-Source: APXvYqwulhvDDmJy/RLd4opW6/BtxntvVuxWuvAOUb7r2tGTe46LVWtACGWsXPDDBsbvTQipeo6wow== X-Received: by 2002:a1c:9d4a:: with SMTP id g71mr4234912wme.50.1575537899000; Thu, 05 Dec 2019 01:24:59 -0800 (PST) Received: from bender.baylibre.local (wal59-h01-176-150-251-154.dsl.sta.abo.bbox.fr. [176.150.251.154]) by smtp.gmail.com with ESMTPSA id t5sm11642620wrr.35.2019.12.05.01.24.57 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Thu, 05 Dec 2019 01:24:58 -0800 (PST) From: Neil Armstrong To: mchehab@kernel.org, hans.verkuil@cisco.com Cc: Maxime Jourdan , linux-media@vger.kernel.org, linux-amlogic@lists.infradead.org, linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org, Neil Armstrong Subject: [PATCH 2/5] media: meson: vdec: add helpers for lossless framebuffer compression buffers Date: Thu, 5 Dec 2019 10:24:51 +0100 Message-Id: <20191205092454.26075-3-narmstrong@baylibre.com> X-Mailer: git-send-email 2.22.0 In-Reply-To: <20191205092454.26075-1-narmstrong@baylibre.com> References: <20191205092454.26075-1-narmstrong@baylibre.com> MIME-Version: 1.0 Sender: linux-media-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-media@vger.kernel.org From: Maxime Jourdan Add helpers to support the lossless framebuffer compression format that will be used in HEVC & VP9 decoders when decoding 10bit content for downsampling to 8bit NV12 and later proper compressed buffer support. Signed-off-by: Maxime Jourdan Signed-off-by: Neil Armstrong --- .../staging/media/meson/vdec/vdec_helpers.c | 27 +++++++++++++++++++ .../staging/media/meson/vdec/vdec_helpers.h | 4 +++ 2 files changed, 31 insertions(+) diff --git a/drivers/staging/media/meson/vdec/vdec_helpers.c b/drivers/staging/media/meson/vdec/vdec_helpers.c index fc59d8801643..818064b6b4d0 100644 --- a/drivers/staging/media/meson/vdec/vdec_helpers.c +++ b/drivers/staging/media/meson/vdec/vdec_helpers.c @@ -50,6 +50,33 @@ void amvdec_write_parser(struct amvdec_core *core, u32 reg, u32 val) } EXPORT_SYMBOL_GPL(amvdec_write_parser); +/* 4 KiB per 64x32 block */ +u32 amvdec_am21c_body_size(u32 width, u32 height) +{ + u32 width_64 = ALIGN(width, 64) / 64; + u32 height_32 = ALIGN(height, 32) / 32; + + return SZ_4K * width_64 * height_32; +} +EXPORT_SYMBOL_GPL(amvdec_am21c_body_size); + +/* 32 bytes per 128x64 block */ +u32 amvdec_am21c_head_size(u32 width, u32 height) +{ + u32 width_128 = ALIGN(width, 128) / 128; + u32 height_64 = ALIGN(height, 64) / 64; + + return 32 * width_128 * height_64; +} +EXPORT_SYMBOL_GPL(amvdec_am21c_head_size); + +u32 amvdec_am21c_size(u32 width, u32 height) +{ + return ALIGN(amvdec_am21c_body_size(width, height) + + amvdec_am21c_head_size(width, height), SZ_64K); +} +EXPORT_SYMBOL_GPL(amvdec_am21c_size); + static int canvas_alloc(struct amvdec_session *sess, u8 *canvas_id) { int ret; diff --git a/drivers/staging/media/meson/vdec/vdec_helpers.h b/drivers/staging/media/meson/vdec/vdec_helpers.h index 165e6293ffba..cfaed52ab526 100644 --- a/drivers/staging/media/meson/vdec/vdec_helpers.h +++ b/drivers/staging/media/meson/vdec/vdec_helpers.h @@ -27,6 +27,10 @@ void amvdec_clear_dos_bits(struct amvdec_core *core, u32 reg, u32 val); u32 amvdec_read_parser(struct amvdec_core *core, u32 reg); void amvdec_write_parser(struct amvdec_core *core, u32 reg, u32 val); +u32 amvdec_am21c_body_size(u32 width, u32 height); +u32 amvdec_am21c_head_size(u32 width, u32 height); +u32 amvdec_am21c_size(u32 width, u32 height); + /** * amvdec_dst_buf_done_idx() - Signal that a buffer is done decoding * From patchwork Thu Dec 5 09:24:52 2019 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Neil Armstrong X-Patchwork-Id: 11274389 Return-Path: Received: from mail.kernel.org (pdx-korg-mail-1.web.codeaurora.org [172.30.200.123]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id 2DC8414B7 for ; Thu, 5 Dec 2019 09:25:10 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id DF01024652 for ; Thu, 5 Dec 2019 09:25:09 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=pass (2048-bit key) header.d=baylibre-com.20150623.gappssmtp.com header.i=@baylibre-com.20150623.gappssmtp.com header.b="TWekmxzr" Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1729179AbfLEJZG (ORCPT ); Thu, 5 Dec 2019 04:25:06 -0500 Received: from mail-wr1-f67.google.com ([209.85.221.67]:39602 "EHLO mail-wr1-f67.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1729147AbfLEJZE (ORCPT ); Thu, 5 Dec 2019 04:25:04 -0500 Received: by mail-wr1-f67.google.com with SMTP id y11so2576310wrt.6 for ; Thu, 05 Dec 2019 01:25:01 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=baylibre-com.20150623.gappssmtp.com; s=20150623; h=from:to:cc:subject:date:message-id:in-reply-to:references :mime-version:content-transfer-encoding; bh=BguqgoE1jCQbIijD1jHMDagjNNEXrcQaeP9djSbm+K4=; b=TWekmxzrQeRV/irBj0EavDZy+dAaeaWIbB4fuGYQcjhG9ekPGgJ7CWiQ0FOiQTxlLZ JL3yQ5tyYpn8qPwX38SSBpnUz9iMudHkwptgd9h6kndxVIjMxaQV0UIevVUD45sNoxcC Brcmrf1LRRLtYpDrZ3W/bX/zaJ5CSX+D5VmN0z+x8UpcVlhFd6eXxrTDdpAs1Cr4G4zn XvLzlOiEYtjCIM96r7Yu7Cu1DqEGlk7Frk8Yc7UT/bNM3jSRTM1J+xdqDrNBDhNfSYg/ UAcCYNbBpE4b60W3leVp3sA3S1y5YVUDmeYVJNByjOIpnEKJ5tDIgdXkzuCJoB0OVnN8 weFw== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to :references:mime-version:content-transfer-encoding; bh=BguqgoE1jCQbIijD1jHMDagjNNEXrcQaeP9djSbm+K4=; b=f8Te5rwa+IbpvxSiPskN7AedR5TOEkO/EHvijMGZD0ZmQv9ujoBvm0Nw+LDVKryMPW PCK6WvjYOEOHWYMbplSMPooztRic2OMpFXqZw6faqFAELuMM+C6ieW9rCa1FESoODwgn ko68/xkSq9SViJFjINViYnS/+K47yXdzv9BoPwsS0vc89lRvTm+aZ8fZtZ9MjkyifMkN NWSmpXOoTaGjwbttZkwN50fFCthRXqTsAN+dRsr9cmfFy10fOeCdaHwRwaZbnazXpEk/ sc0u1e2M3xCrohTaSX8rw9NfFUNRXD8Po91IbBFwpUWzyoUwHAjLsq4Ac9ecYE8az5Q0 38Qg== X-Gm-Message-State: APjAAAXlCOkYS9JCwKp6ZLckGYSF3YTrme5uFmnJ95rwDAaLpVjyBIbj YeT869Dm10ejFSeDmcaBAoInag== X-Google-Smtp-Source: APXvYqxTHnaKe01tt1DSQDh0/LDUqVgzhSAD+cjmQ5akcBD44ToaMjUolrqSJHOVnkCw9I8ntxWDtQ== X-Received: by 2002:a5d:55d1:: with SMTP id i17mr8690191wrw.165.1575537900209; Thu, 05 Dec 2019 01:25:00 -0800 (PST) Received: from bender.baylibre.local (wal59-h01-176-150-251-154.dsl.sta.abo.bbox.fr. [176.150.251.154]) by smtp.gmail.com with ESMTPSA id t5sm11642620wrr.35.2019.12.05.01.24.59 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Thu, 05 Dec 2019 01:24:59 -0800 (PST) From: Neil Armstrong To: mchehab@kernel.org, hans.verkuil@cisco.com Cc: Maxime Jourdan , linux-media@vger.kernel.org, linux-amlogic@lists.infradead.org, linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org, Neil Armstrong Subject: [PATCH 3/5] media: meson: vdec: add common HEVC decoder support Date: Thu, 5 Dec 2019 10:24:52 +0100 Message-Id: <20191205092454.26075-4-narmstrong@baylibre.com> X-Mailer: git-send-email 2.22.0 In-Reply-To: <20191205092454.26075-1-narmstrong@baylibre.com> References: <20191205092454.26075-1-narmstrong@baylibre.com> MIME-Version: 1.0 Sender: linux-media-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-media@vger.kernel.org From: Maxime Jourdan Add support for the HEVC & VP9 common decoder support, handling Amlogic GXBB, GXL, G12A and SM1 platforms. This handles the "HEVC" hw decoder used for HEVC and VP9, and will be using in the new H264 multi-instance decoder for G12A & SM1 platforms. Signed-off-by: Maxime Jourdan Signed-off-by: Neil Armstrong --- drivers/staging/media/meson/vdec/Makefile | 4 +- .../media/meson/vdec/codec_hevc_common.c | 285 ++++++++++++++++++ .../media/meson/vdec/codec_hevc_common.h | 77 +++++ drivers/staging/media/meson/vdec/hevc_regs.h | 211 +++++++++++++ drivers/staging/media/meson/vdec/vdec_hevc.c | 231 ++++++++++++++ drivers/staging/media/meson/vdec/vdec_hevc.h | 13 + 6 files changed, 819 insertions(+), 2 deletions(-) create mode 100644 drivers/staging/media/meson/vdec/codec_hevc_common.c create mode 100644 drivers/staging/media/meson/vdec/codec_hevc_common.h create mode 100644 drivers/staging/media/meson/vdec/hevc_regs.h create mode 100644 drivers/staging/media/meson/vdec/vdec_hevc.c create mode 100644 drivers/staging/media/meson/vdec/vdec_hevc.h diff --git a/drivers/staging/media/meson/vdec/Makefile b/drivers/staging/media/meson/vdec/Makefile index 711d990c760e..f55b6e625034 100644 --- a/drivers/staging/media/meson/vdec/Makefile +++ b/drivers/staging/media/meson/vdec/Makefile @@ -2,7 +2,7 @@ # Makefile for Amlogic meson video decoder driver meson-vdec-objs = esparser.o vdec.o vdec_helpers.o vdec_platform.o -meson-vdec-objs += vdec_1.o -meson-vdec-objs += codec_mpeg12.o codec_h264.o +meson-vdec-objs += vdec_1.o vdec_hevc.o +meson-vdec-objs += codec_mpeg12.o codec_h264.o codec_hevc_common.o obj-$(CONFIG_VIDEO_MESON_VDEC) += meson-vdec.o diff --git a/drivers/staging/media/meson/vdec/codec_hevc_common.c b/drivers/staging/media/meson/vdec/codec_hevc_common.c new file mode 100644 index 000000000000..72a4e38a42dc --- /dev/null +++ b/drivers/staging/media/meson/vdec/codec_hevc_common.c @@ -0,0 +1,285 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* + * Copyright (C) 2018 Maxime Jourdan + */ + +#include +#include + +#include "codec_hevc_common.h" +#include "vdec_helpers.h" +#include "hevc_regs.h" + +#define MMU_COMPRESS_HEADER_SIZE 0x48000 +#define MMU_MAP_SIZE 0x4800 + +/* Configure decode head read mode */ +void codec_hevc_setup_decode_head(struct amvdec_session *sess, int is_10bit) +{ + struct amvdec_core *core = sess->core; + u32 body_size = amvdec_am21c_body_size(sess->width, sess->height); + u32 head_size = amvdec_am21c_head_size(sess->width, sess->height); + + if (!codec_hevc_use_fbc(sess->pixfmt_cap, is_10bit)) { + /* Enable 2-plane reference read mode */ + amvdec_write_dos(core, HEVCD_MPP_DECOMP_CTL1, BIT(31)); + return; + } + + if (codec_hevc_use_mmu(core->platform->revision, + sess->pixfmt_cap, is_10bit)) + amvdec_write_dos(core, HEVCD_MPP_DECOMP_CTL1, BIT(4)); + else + amvdec_write_dos(core, HEVCD_MPP_DECOMP_CTL1, 0); + + amvdec_write_dos(core, HEVCD_MPP_DECOMP_CTL2, body_size / 32); + amvdec_write_dos(core, HEVC_CM_BODY_LENGTH, body_size); + amvdec_write_dos(core, HEVC_CM_HEADER_OFFSET, body_size); + amvdec_write_dos(core, HEVC_CM_HEADER_LENGTH, head_size); +} +EXPORT_SYMBOL_GPL(codec_hevc_setup_decode_head); + +static void codec_hevc_setup_buffers_gxbb(struct amvdec_session *sess, + struct codec_hevc_common *comm, + int is_10bit) +{ + struct amvdec_core *core = sess->core; + struct v4l2_m2m_buffer *buf; + u32 buf_num = v4l2_m2m_num_dst_bufs_ready(sess->m2m_ctx); + dma_addr_t buf_y_paddr = 0; + dma_addr_t buf_uv_paddr = 0; + u32 idx = 0; + u32 val; + int i; + + amvdec_write_dos(core, HEVCD_MPP_ANC2AXI_TBL_CONF_ADDR, 0); + + v4l2_m2m_for_each_dst_buf(sess->m2m_ctx, buf) { + struct vb2_buffer *vb = &buf->vb.vb2_buf; + + idx = vb->index; + + if (codec_hevc_use_downsample(sess->pixfmt_cap, is_10bit)) + buf_y_paddr = comm->fbc_buffer_paddr[idx]; + else + buf_y_paddr = vb2_dma_contig_plane_dma_addr(vb, 0); + + if (codec_hevc_use_fbc(sess->pixfmt_cap, is_10bit)) { + val = buf_y_paddr | (idx << 8) | 1; + amvdec_write_dos(core, HEVCD_MPP_ANC2AXI_TBL_CMD_ADDR, + val); + } else { + buf_uv_paddr = vb2_dma_contig_plane_dma_addr(vb, 1); + val = buf_y_paddr | ((idx * 2) << 8) | 1; + amvdec_write_dos(core, HEVCD_MPP_ANC2AXI_TBL_CMD_ADDR, + val); + val = buf_uv_paddr | ((idx * 2 + 1) << 8) | 1; + amvdec_write_dos(core, HEVCD_MPP_ANC2AXI_TBL_CMD_ADDR, + val); + } + } + + if (codec_hevc_use_fbc(sess->pixfmt_cap, is_10bit)) + val = buf_y_paddr | (idx << 8) | 1; + else + val = buf_y_paddr | ((idx * 2) << 8) | 1; + + /* Fill the remaining unused slots with the last buffer's Y addr */ + for (i = buf_num; i < MAX_REF_PIC_NUM; ++i) + amvdec_write_dos(core, HEVCD_MPP_ANC2AXI_TBL_CMD_ADDR, val); + + amvdec_write_dos(core, HEVCD_MPP_ANC2AXI_TBL_CONF_ADDR, 1); + amvdec_write_dos(core, HEVCD_MPP_ANC_CANVAS_ACCCONFIG_ADDR, 1); + for (i = 0; i < 32; ++i) + amvdec_write_dos(core, HEVCD_MPP_ANC_CANVAS_DATA_ADDR, 0); +} + +static void codec_hevc_setup_buffers_gxl(struct amvdec_session *sess, + struct codec_hevc_common *comm, + int is_10bit) +{ + struct amvdec_core *core = sess->core; + struct v4l2_m2m_buffer *buf; + u32 revision = core->platform->revision; + u32 pixfmt_cap = sess->pixfmt_cap; + int i; + + amvdec_write_dos(core, HEVCD_MPP_ANC2AXI_TBL_CONF_ADDR, + BIT(2) | BIT(1)); + + v4l2_m2m_for_each_dst_buf(sess->m2m_ctx, buf) { + struct vb2_buffer *vb = &buf->vb.vb2_buf; + dma_addr_t buf_y_paddr = 0; + dma_addr_t buf_uv_paddr = 0; + u32 idx = vb->index; + + if (codec_hevc_use_mmu(revision, pixfmt_cap, is_10bit)) + buf_y_paddr = comm->mmu_header_paddr[idx]; + else if (codec_hevc_use_downsample(pixfmt_cap, is_10bit)) + buf_y_paddr = comm->fbc_buffer_paddr[idx]; + else + buf_y_paddr = vb2_dma_contig_plane_dma_addr(vb, 0); + + amvdec_write_dos(core, HEVCD_MPP_ANC2AXI_TBL_DATA, + buf_y_paddr >> 5); + + if (!codec_hevc_use_fbc(pixfmt_cap, is_10bit)) { + buf_uv_paddr = vb2_dma_contig_plane_dma_addr(vb, 1); + amvdec_write_dos(core, HEVCD_MPP_ANC2AXI_TBL_DATA, + buf_uv_paddr >> 5); + } + } + + amvdec_write_dos(core, HEVCD_MPP_ANC2AXI_TBL_CONF_ADDR, 1); + amvdec_write_dos(core, HEVCD_MPP_ANC_CANVAS_ACCCONFIG_ADDR, 1); + for (i = 0; i < 32; ++i) + amvdec_write_dos(core, HEVCD_MPP_ANC_CANVAS_DATA_ADDR, 0); +} + +void codec_hevc_free_fbc_buffers(struct amvdec_session *sess, + struct codec_hevc_common *comm) +{ + struct device *dev = sess->core->dev; + u32 am21_size = amvdec_am21c_size(sess->width, sess->height); + int i; + + for (i = 0; i < MAX_REF_PIC_NUM; ++i) { + if (comm->fbc_buffer_vaddr[i]) { + dma_free_coherent(dev, am21_size, + comm->fbc_buffer_vaddr[i], + comm->fbc_buffer_paddr[i]); + comm->fbc_buffer_vaddr[i] = NULL; + } + } +} +EXPORT_SYMBOL_GPL(codec_hevc_free_fbc_buffers); + +static int codec_hevc_alloc_fbc_buffers(struct amvdec_session *sess, + struct codec_hevc_common *comm) +{ + struct device *dev = sess->core->dev; + struct v4l2_m2m_buffer *buf; + u32 am21_size = amvdec_am21c_size(sess->width, sess->height); + + v4l2_m2m_for_each_dst_buf(sess->m2m_ctx, buf) { + u32 idx = buf->vb.vb2_buf.index; + dma_addr_t paddr; + void *vaddr = dma_alloc_coherent(dev, am21_size, &paddr, + GFP_KERNEL); + if (!vaddr) { + dev_err(dev, "Couldn't allocate FBC buffer %u\n", idx); + codec_hevc_free_fbc_buffers(sess, comm); + return -ENOMEM; + } + + comm->fbc_buffer_vaddr[idx] = vaddr; + comm->fbc_buffer_paddr[idx] = paddr; + } + + return 0; +} + +void codec_hevc_free_mmu_headers(struct amvdec_session *sess, + struct codec_hevc_common *comm) +{ + struct device *dev = sess->core->dev; + int i; + + for (i = 0; i < MAX_REF_PIC_NUM; ++i) { + if (comm->mmu_header_vaddr[i]) { + dma_free_coherent(dev, MMU_COMPRESS_HEADER_SIZE, + comm->mmu_header_vaddr[i], + comm->mmu_header_paddr[i]); + comm->mmu_header_vaddr[i] = NULL; + } + } + + if (comm->mmu_map_vaddr) { + dma_free_coherent(dev, MMU_MAP_SIZE, + comm->mmu_map_vaddr, + comm->mmu_map_paddr); + comm->mmu_map_vaddr = NULL; + } +} +EXPORT_SYMBOL_GPL(codec_hevc_free_mmu_headers); + +static int codec_hevc_alloc_mmu_headers(struct amvdec_session *sess, + struct codec_hevc_common *comm) +{ + struct device *dev = sess->core->dev; + struct v4l2_m2m_buffer *buf; + + comm->mmu_map_vaddr = dma_alloc_coherent(dev, MMU_MAP_SIZE, + &comm->mmu_map_paddr, + GFP_KERNEL); + if (!comm->mmu_map_vaddr) + return -ENOMEM; + + v4l2_m2m_for_each_dst_buf(sess->m2m_ctx, buf) { + u32 idx = buf->vb.vb2_buf.index; + dma_addr_t paddr; + void *vaddr = dma_alloc_coherent(dev, MMU_COMPRESS_HEADER_SIZE, + &paddr, GFP_KERNEL); + if (!vaddr) { + dev_err(dev, "Couldn't allocate MMU header %u\n", idx); + codec_hevc_free_mmu_headers(sess, comm); + return -ENOMEM; + } + + comm->mmu_header_vaddr[idx] = vaddr; + comm->mmu_header_paddr[idx] = paddr; + } + + return 0; +} + +int codec_hevc_setup_buffers(struct amvdec_session *sess, + struct codec_hevc_common *comm, + int is_10bit) +{ + struct amvdec_core *core = sess->core; + int ret; + + if (codec_hevc_use_downsample(sess->pixfmt_cap, is_10bit)) { + ret = codec_hevc_alloc_fbc_buffers(sess, comm); + if (ret) + return ret; + } + + if (codec_hevc_use_mmu(core->platform->revision, + sess->pixfmt_cap, is_10bit)) { + ret = codec_hevc_alloc_mmu_headers(sess, comm); + if (ret) { + codec_hevc_free_fbc_buffers(sess, comm); + return ret; + } + } + + if (core->platform->revision == VDEC_REVISION_GXBB) + codec_hevc_setup_buffers_gxbb(sess, comm, is_10bit); + else + codec_hevc_setup_buffers_gxl(sess, comm, is_10bit); + + return 0; +} +EXPORT_SYMBOL_GPL(codec_hevc_setup_buffers); + +void codec_hevc_fill_mmu_map(struct amvdec_session *sess, + struct codec_hevc_common *comm, + struct vb2_buffer *vb) +{ + u32 size = amvdec_am21c_size(sess->width, sess->height); + u32 nb_pages = size / PAGE_SIZE; + u32 *mmu_map = comm->mmu_map_vaddr; + u32 first_page; + u32 i; + + if (sess->pixfmt_cap == V4L2_PIX_FMT_NV12M) + first_page = comm->fbc_buffer_paddr[vb->index] >> PAGE_SHIFT; + else + first_page = vb2_dma_contig_plane_dma_addr(vb, 0) >> PAGE_SHIFT; + + for (i = 0; i < nb_pages; ++i) + mmu_map[i] = first_page + i; +} +EXPORT_SYMBOL_GPL(codec_hevc_fill_mmu_map); diff --git a/drivers/staging/media/meson/vdec/codec_hevc_common.h b/drivers/staging/media/meson/vdec/codec_hevc_common.h new file mode 100644 index 000000000000..de16d2e43061 --- /dev/null +++ b/drivers/staging/media/meson/vdec/codec_hevc_common.h @@ -0,0 +1,77 @@ +/* SPDX-License-Identifier: GPL-2.0+ */ +/* + * Copyright (C) 2018 BayLibre, SAS + * Author: Maxime Jourdan + */ + +#ifndef __MESON_VDEC_HEVC_COMMON_H_ +#define __MESON_VDEC_HEVC_COMMON_H_ + +#include "vdec.h" + +#define PARSER_CMD_SKIP_CFG_0 0x0000090b +#define PARSER_CMD_SKIP_CFG_1 0x1b14140f +#define PARSER_CMD_SKIP_CFG_2 0x001b1910 +static const u16 vdec_hevc_parser_cmd[] = { + 0x0401, 0x8401, 0x0800, 0x0402, + 0x9002, 0x1423, 0x8CC3, 0x1423, + 0x8804, 0x9825, 0x0800, 0x04FE, + 0x8406, 0x8411, 0x1800, 0x8408, + 0x8409, 0x8C2A, 0x9C2B, 0x1C00, + 0x840F, 0x8407, 0x8000, 0x8408, + 0x2000, 0xA800, 0x8410, 0x04DE, + 0x840C, 0x840D, 0xAC00, 0xA000, + 0x08C0, 0x08E0, 0xA40E, 0xFC00, + 0x7C00 +}; + +#define MAX_REF_PIC_NUM 24 + +struct codec_hevc_common { + void *fbc_buffer_vaddr[MAX_REF_PIC_NUM]; + dma_addr_t fbc_buffer_paddr[MAX_REF_PIC_NUM]; + + void *mmu_header_vaddr[MAX_REF_PIC_NUM]; + dma_addr_t mmu_header_paddr[MAX_REF_PIC_NUM]; + + void *mmu_map_vaddr; + dma_addr_t mmu_map_paddr; +}; + +/* Returns 1 if we must use framebuffer compression */ +static inline int codec_hevc_use_fbc(u32 pixfmt, int is_10bit) +{ + /* TOFIX: Handle Amlogic Compressed buffer for 8bit also */ + return is_10bit; +} + +/* Returns 1 if we are decoding 10-bit but outputting 8-bit NV12 */ +static inline int codec_hevc_use_downsample(u32 pixfmt, int is_10bit) +{ + return is_10bit; +} + +/* Returns 1 if we are decoding using the IOMMU */ +static inline int codec_hevc_use_mmu(u32 revision, u32 pixfmt, int is_10bit) +{ + return revision >= VDEC_REVISION_G12A && + codec_hevc_use_fbc(pixfmt, is_10bit); +} + +/** + * Configure decode head read mode + */ +void codec_hevc_setup_decode_head(struct amvdec_session *sess, int is_10bit); + +void codec_hevc_free_fbc_buffers(struct amvdec_session *sess, + struct codec_hevc_common *comm); + +int codec_hevc_setup_buffers(struct amvdec_session *sess, + struct codec_hevc_common *comm, + int is_10bit); + +void codec_hevc_fill_mmu_map(struct amvdec_session *sess, + struct codec_hevc_common *comm, + struct vb2_buffer *vb); + +#endif diff --git a/drivers/staging/media/meson/vdec/hevc_regs.h b/drivers/staging/media/meson/vdec/hevc_regs.h new file mode 100644 index 000000000000..55c1a80b955a --- /dev/null +++ b/drivers/staging/media/meson/vdec/hevc_regs.h @@ -0,0 +1,211 @@ +/* SPDX-License-Identifier: GPL-2.0+ */ +/* + * Copyright (C) 2015 Amlogic, Inc. All rights reserved. + */ + +#ifndef __MESON_VDEC_HEVC_REGS_H_ +#define __MESON_VDEC_HEVC_REGS_H_ + +#define HEVC_ASSIST_MMU_MAP_ADDR 0xc024 + +#define HEVC_ASSIST_MBOX1_CLR_REG 0xc1d4 +#define HEVC_ASSIST_MBOX1_MASK 0xc1d8 + +#define HEVC_ASSIST_SCRATCH_0 0xc300 +#define HEVC_ASSIST_SCRATCH_1 0xc304 +#define HEVC_ASSIST_SCRATCH_2 0xc308 +#define HEVC_ASSIST_SCRATCH_3 0xc30c +#define HEVC_ASSIST_SCRATCH_4 0xc310 +#define HEVC_ASSIST_SCRATCH_5 0xc314 +#define HEVC_ASSIST_SCRATCH_6 0xc318 +#define HEVC_ASSIST_SCRATCH_7 0xc31c +#define HEVC_ASSIST_SCRATCH_8 0xc320 +#define HEVC_ASSIST_SCRATCH_9 0xc324 +#define HEVC_ASSIST_SCRATCH_A 0xc328 +#define HEVC_ASSIST_SCRATCH_B 0xc32c +#define HEVC_ASSIST_SCRATCH_C 0xc330 +#define HEVC_ASSIST_SCRATCH_D 0xc334 +#define HEVC_ASSIST_SCRATCH_E 0xc338 +#define HEVC_ASSIST_SCRATCH_F 0xc33c +#define HEVC_ASSIST_SCRATCH_G 0xc340 +#define HEVC_ASSIST_SCRATCH_H 0xc344 +#define HEVC_ASSIST_SCRATCH_I 0xc348 +#define HEVC_ASSIST_SCRATCH_J 0xc34c +#define HEVC_ASSIST_SCRATCH_K 0xc350 +#define HEVC_ASSIST_SCRATCH_L 0xc354 +#define HEVC_ASSIST_SCRATCH_M 0xc358 +#define HEVC_ASSIST_SCRATCH_N 0xc35c + +#define HEVC_PARSER_VERSION 0xc400 +#define HEVC_STREAM_CONTROL 0xc404 +#define HEVC_STREAM_START_ADDR 0xc408 +#define HEVC_STREAM_END_ADDR 0xc40c +#define HEVC_STREAM_WR_PTR 0xc410 +#define HEVC_STREAM_RD_PTR 0xc414 +#define HEVC_STREAM_LEVEL 0xc418 +#define HEVC_STREAM_FIFO_CTL 0xc41c +#define HEVC_SHIFT_CONTROL 0xc420 +#define HEVC_SHIFT_STARTCODE 0xc424 +#define HEVC_SHIFT_EMULATECODE 0xc428 +#define HEVC_SHIFT_STATUS 0xc42c +#define HEVC_SHIFTED_DATA 0xc430 +#define HEVC_SHIFT_BYTE_COUNT 0xc434 +#define HEVC_SHIFT_COMMAND 0xc438 +#define HEVC_ELEMENT_RESULT 0xc43c +#define HEVC_CABAC_CONTROL 0xc440 +#define HEVC_PARSER_SLICE_INFO 0xc444 +#define HEVC_PARSER_CMD_WRITE 0xc448 +#define HEVC_PARSER_CORE_CONTROL 0xc44c +#define HEVC_PARSER_CMD_FETCH 0xc450 +#define HEVC_PARSER_CMD_STATUS 0xc454 +#define HEVC_PARSER_LCU_INFO 0xc458 +#define HEVC_PARSER_HEADER_INFO 0xc45c +#define HEVC_PARSER_INT_CONTROL 0xc480 +#define HEVC_PARSER_INT_STATUS 0xc484 +#define HEVC_PARSER_IF_CONTROL 0xc488 +#define HEVC_PARSER_PICTURE_SIZE 0xc48c +#define HEVC_PARSER_LCU_START 0xc490 +#define HEVC_PARSER_HEADER_INFO2 0xc494 +#define HEVC_PARSER_QUANT_READ 0xc498 +#define HEVC_PARSER_RESERVED_27 0xc49c +#define HEVC_PARSER_CMD_SKIP_0 0xc4a0 +#define HEVC_PARSER_CMD_SKIP_1 0xc4a4 +#define HEVC_PARSER_CMD_SKIP_2 0xc4a8 +#define HEVC_SAO_IF_STATUS 0xc4c0 +#define HEVC_SAO_IF_DATA_Y 0xc4c4 +#define HEVC_SAO_IF_DATA_U 0xc4c8 +#define HEVC_SAO_IF_DATA_V 0xc4cc +#define HEVC_STREAM_SWAP_ADDR 0xc4d0 +#define HEVC_STREAM_SWAP_CTRL 0xc4d4 +#define HEVC_IQIT_IF_WAIT_CNT 0xc4d8 +#define HEVC_MPRED_IF_WAIT_CNT 0xc4dc +#define HEVC_SAO_IF_WAIT_CNT 0xc4e0 + +#define HEVC_MPRED_VERSION 0xc800 +#define HEVC_MPRED_CTRL0 0xc804 + #define MPRED_CTRL0_NEW_PIC BIT(2) + #define MPRED_CTRL0_NEW_TILE BIT(3) + #define MPRED_CTRL0_NEW_SLI_SEG BIT(4) + #define MPRED_CTRL0_TMVP BIT(5) + #define MPRED_CTRL0_LDC BIT(6) + #define MPRED_CTRL0_COL_FROM_L0 BIT(7) + #define MPRED_CTRL0_ABOVE_EN BIT(9) + #define MPRED_CTRL0_MV_WR_EN BIT(10) + #define MPRED_CTRL0_MV_RD_EN BIT(11) + #define MPRED_CTRL0_BUF_LINEAR BIT(13) +#define HEVC_MPRED_CTRL1 0xc808 +#define HEVC_MPRED_INT_EN 0xc80c +#define HEVC_MPRED_INT_STATUS 0xc810 +#define HEVC_MPRED_PIC_SIZE 0xc814 +#define HEVC_MPRED_PIC_SIZE_LCU 0xc818 +#define HEVC_MPRED_TILE_START 0xc81c +#define HEVC_MPRED_TILE_SIZE_LCU 0xc820 +#define HEVC_MPRED_REF_NUM 0xc824 +#define HEVC_MPRED_REF_EN_L0 0xc830 +#define HEVC_MPRED_REF_EN_L1 0xc834 +#define HEVC_MPRED_COLREF_EN_L0 0xc838 +#define HEVC_MPRED_COLREF_EN_L1 0xc83c +#define HEVC_MPRED_AXI_WCTRL 0xc840 +#define HEVC_MPRED_AXI_RCTRL 0xc844 +#define HEVC_MPRED_ABV_START_ADDR 0xc848 +#define HEVC_MPRED_MV_WR_START_ADDR 0xc84c +#define HEVC_MPRED_MV_RD_START_ADDR 0xc850 +#define HEVC_MPRED_MV_WPTR 0xc854 +#define HEVC_MPRED_MV_RPTR 0xc858 +#define HEVC_MPRED_MV_WR_ROW_JUMP 0xc85c +#define HEVC_MPRED_MV_RD_ROW_JUMP 0xc860 +#define HEVC_MPRED_CURR_LCU 0xc864 +#define HEVC_MPRED_ABV_WPTR 0xc868 +#define HEVC_MPRED_ABV_RPTR 0xc86c +#define HEVC_MPRED_CTRL2 0xc870 +#define HEVC_MPRED_CTRL3 0xc874 +#define HEVC_MPRED_L0_REF00_POC 0xc880 +#define HEVC_MPRED_L1_REF00_POC 0xc8c0 + +#define HEVC_MPRED_CUR_POC 0xc980 +#define HEVC_MPRED_COL_POC 0xc984 +#define HEVC_MPRED_MV_RD_END_ADDR 0xc988 + +#define HEVC_MSP 0xcc00 +#define HEVC_MPSR 0xcc04 +#define HEVC_MCPU_INTR_MSK 0xcc10 +#define HEVC_MCPU_INTR_REQ 0xcc14 +#define HEVC_CPSR 0xcc84 + +#define HEVC_IMEM_DMA_CTRL 0xcd00 +#define HEVC_IMEM_DMA_ADR 0xcd04 +#define HEVC_IMEM_DMA_COUNT 0xcd08 + +#define HEVCD_IPP_TOP_CNTL 0xd000 +#define HEVCD_IPP_LINEBUFF_BASE 0xd024 +#define HEVCD_IPP_AXIIF_CONFIG 0xd02c + +#define HEVCD_MPP_ANC2AXI_TBL_CONF_ADDR 0xd180 +#define HEVCD_MPP_ANC2AXI_TBL_CMD_ADDR 0xd184 +#define HEVCD_MPP_ANC2AXI_TBL_DATA 0xd190 + +#define HEVCD_MPP_ANC_CANVAS_ACCCONFIG_ADDR 0xd300 +#define HEVCD_MPP_ANC_CANVAS_DATA_ADDR 0xd304 +#define HEVCD_MPP_DECOMP_CTL1 0xd308 +#define HEVCD_MPP_DECOMP_CTL2 0xd30c +#define HEVCD_MCRCC_CTL1 0xd3c0 +#define HEVCD_MCRCC_CTL2 0xd3c4 +#define HEVCD_MCRCC_CTL3 0xd3c8 + +#define HEVC_DBLK_CFG0 0xd400 +#define HEVC_DBLK_CFG1 0xd404 +#define HEVC_DBLK_CFG2 0xd408 +#define HEVC_DBLK_CFG3 0xd40c +#define HEVC_DBLK_CFG4 0xd410 +#define HEVC_DBLK_CFG5 0xd414 +#define HEVC_DBLK_CFG6 0xd418 +#define HEVC_DBLK_CFG7 0xd41c +#define HEVC_DBLK_CFG8 0xd420 +#define HEVC_DBLK_CFG9 0xd424 +#define HEVC_DBLK_CFGA 0xd428 +#define HEVC_DBLK_STS0 0xd42c +#define HEVC_DBLK_STS1 0xd430 +#define HEVC_DBLK_CFGE 0xd438 + +#define HEVC_SAO_VERSION 0xd800 +#define HEVC_SAO_CTRL0 0xd804 +#define HEVC_SAO_CTRL1 0xd808 +#define HEVC_SAO_PIC_SIZE 0xd814 +#define HEVC_SAO_PIC_SIZE_LCU 0xd818 +#define HEVC_SAO_TILE_START 0xd81c +#define HEVC_SAO_TILE_SIZE_LCU 0xd820 +#define HEVC_SAO_Y_START_ADDR 0xd82c +#define HEVC_SAO_Y_LENGTH 0xd830 +#define HEVC_SAO_C_START_ADDR 0xd834 +#define HEVC_SAO_C_LENGTH 0xd838 +#define HEVC_SAO_Y_WPTR 0xd83c +#define HEVC_SAO_C_WPTR 0xd840 +#define HEVC_SAO_ABV_START_ADDR 0xd844 +#define HEVC_SAO_VB_WR_START_ADDR 0xd848 +#define HEVC_SAO_VB_RD_START_ADDR 0xd84c +#define HEVC_SAO_ABV_WPTR 0xd850 +#define HEVC_SAO_ABV_RPTR 0xd854 +#define HEVC_SAO_VB_WPTR 0xd858 +#define HEVC_SAO_VB_RPTR 0xd85c +#define HEVC_SAO_CTRL2 0xd880 +#define HEVC_SAO_CTRL3 0xd884 +#define HEVC_SAO_CTRL4 0xd888 +#define HEVC_SAO_CTRL5 0xd88c +#define HEVC_SAO_CTRL6 0xd890 +#define HEVC_SAO_CTRL7 0xd894 +#define HEVC_CM_BODY_START_ADDR 0xd898 +#define HEVC_CM_BODY_LENGTH 0xd89c +#define HEVC_CM_HEADER_START_ADDR 0xd8a0 +#define HEVC_CM_HEADER_LENGTH 0xd8a4 +#define HEVC_CM_HEADER_OFFSET 0xd8ac +#define HEVC_SAO_MMU_VH0_ADDR 0xd8e8 +#define HEVC_SAO_MMU_VH1_ADDR 0xd8ec + +#define HEVC_IQIT_CLK_RST_CTRL 0xdc00 +#define HEVC_IQIT_SCALELUT_WR_ADDR 0xdc08 +#define HEVC_IQIT_SCALELUT_RD_ADDR 0xdc0c +#define HEVC_IQIT_SCALELUT_DATA 0xdc10 + +#define HEVC_PSCALE_CTRL 0xe444 + +#endif diff --git a/drivers/staging/media/meson/vdec/vdec_hevc.c b/drivers/staging/media/meson/vdec/vdec_hevc.c new file mode 100644 index 000000000000..af41215e106c --- /dev/null +++ b/drivers/staging/media/meson/vdec/vdec_hevc.c @@ -0,0 +1,231 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* + * Copyright (C) 2018 Maxime Jourdan + * + * VDEC_HEVC is a video decoding block that allows decoding of + * HEVC, VP9 + */ + +#include +#include + +#include "vdec_1.h" +#include "vdec_helpers.h" +#include "hevc_regs.h" +#include "dos_regs.h" + +/* AO Registers */ +#define AO_RTI_GEN_PWR_SLEEP0 0xe8 +#define AO_RTI_GEN_PWR_ISO0 0xec + #define GEN_PWR_VDEC_HEVC (BIT(7) | BIT(6)) + #define GEN_PWR_VDEC_HEVC_SM1 (BIT(2)) + +#define MC_SIZE (4096 * 4) + +static int vdec_hevc_load_firmware(struct amvdec_session *sess, + const char *fwname) +{ + struct amvdec_core *core = sess->core; + struct device *dev = core->dev_dec; + const struct firmware *fw; + static void *mc_addr; + static dma_addr_t mc_addr_map; + int ret; + u32 i = 100; + + ret = request_firmware(&fw, fwname, dev); + if (ret < 0) { + dev_err(dev, "Unable to request firmware %s\n", fwname); + return ret; + } + + if (fw->size < MC_SIZE) { + dev_err(dev, "Firmware size %zu is too small. Expected %u.\n", + fw->size, MC_SIZE); + ret = -EINVAL; + goto release_firmware; + } + + mc_addr = dma_alloc_coherent(core->dev, MC_SIZE, &mc_addr_map, + GFP_KERNEL); + if (!mc_addr) { + dev_err(dev, "Failed allocating memory for firmware loading\n"); + ret = -ENOMEM; + goto release_firmware; + } + + memcpy(mc_addr, fw->data, MC_SIZE); + + amvdec_write_dos(core, HEVC_MPSR, 0); + amvdec_write_dos(core, HEVC_CPSR, 0); + + amvdec_write_dos(core, HEVC_IMEM_DMA_ADR, mc_addr_map); + amvdec_write_dos(core, HEVC_IMEM_DMA_COUNT, MC_SIZE / 4); + amvdec_write_dos(core, HEVC_IMEM_DMA_CTRL, (0x8000 | (7 << 16))); + + while (i && (readl(core->dos_base + HEVC_IMEM_DMA_CTRL) & 0x8000)) + i--; + + if (i == 0) { + dev_err(dev, "Firmware load fail (DMA hang?)\n"); + ret = -ENODEV; + } + + dma_free_coherent(core->dev, MC_SIZE, mc_addr, mc_addr_map); +release_firmware: + release_firmware(fw); + return ret; +} + +static void vdec_hevc_stbuf_init(struct amvdec_session *sess) +{ + struct amvdec_core *core = sess->core; + + amvdec_write_dos(core, HEVC_STREAM_CONTROL, + amvdec_read_dos(core, HEVC_STREAM_CONTROL) & ~1); + amvdec_write_dos(core, HEVC_STREAM_START_ADDR, sess->vififo_paddr); + amvdec_write_dos(core, HEVC_STREAM_END_ADDR, + sess->vififo_paddr + sess->vififo_size); + amvdec_write_dos(core, HEVC_STREAM_RD_PTR, sess->vififo_paddr); + amvdec_write_dos(core, HEVC_STREAM_WR_PTR, sess->vififo_paddr); +} + +/* VDEC_HEVC specific ESPARSER configuration */ +static void vdec_hevc_conf_esparser(struct amvdec_session *sess) +{ + struct amvdec_core *core = sess->core; + + /* set vififo_vbuf_rp_sel=>vdec_hevc */ + amvdec_write_dos(core, DOS_GEN_CTRL0, 3 << 1); + amvdec_write_dos(core, HEVC_STREAM_CONTROL, + amvdec_read_dos(core, HEVC_STREAM_CONTROL) | BIT(3)); + amvdec_write_dos(core, HEVC_STREAM_CONTROL, + amvdec_read_dos(core, HEVC_STREAM_CONTROL) | 1); + amvdec_write_dos(core, HEVC_STREAM_FIFO_CTL, + amvdec_read_dos(core, HEVC_STREAM_FIFO_CTL) | BIT(29)); +} + +static u32 vdec_hevc_vififo_level(struct amvdec_session *sess) +{ + return readl_relaxed(sess->core->dos_base + HEVC_STREAM_LEVEL); +} + +static int vdec_hevc_stop(struct amvdec_session *sess) +{ + struct amvdec_core *core = sess->core; + struct amvdec_codec_ops *codec_ops = sess->fmt_out->codec_ops; + + /* Disable interrupt */ + amvdec_write_dos(core, HEVC_ASSIST_MBOX1_MASK, 0); + /* Disable firmware processor */ + amvdec_write_dos(core, HEVC_MPSR, 0); + + if (sess->priv) + codec_ops->stop(sess); + + /* Enable VDEC_HEVC Isolation */ + if (core->platform->revision == VDEC_REVISION_SM1) + regmap_update_bits(core->regmap_ao, AO_RTI_GEN_PWR_ISO0, + GEN_PWR_VDEC_HEVC_SM1, + GEN_PWR_VDEC_HEVC_SM1); + else + regmap_update_bits(core->regmap_ao, AO_RTI_GEN_PWR_ISO0, + 0xc00, 0xc00); + + /* VDEC_HEVC Memories */ + amvdec_write_dos(core, DOS_MEM_PD_HEVC, 0xffffffffUL); + + if (core->platform->revision == VDEC_REVISION_SM1) + regmap_update_bits(core->regmap_ao, AO_RTI_GEN_PWR_SLEEP0, + GEN_PWR_VDEC_HEVC_SM1, + GEN_PWR_VDEC_HEVC_SM1); + else + regmap_update_bits(core->regmap_ao, AO_RTI_GEN_PWR_SLEEP0, + GEN_PWR_VDEC_HEVC, GEN_PWR_VDEC_HEVC); + + clk_disable_unprepare(core->vdec_hevc_clk); + if (core->platform->revision == VDEC_REVISION_G12A || + core->platform->revision == VDEC_REVISION_SM1) + clk_disable_unprepare(core->vdec_hevcf_clk); + + return 0; +} + +static int vdec_hevc_start(struct amvdec_session *sess) +{ + int ret; + struct amvdec_core *core = sess->core; + struct amvdec_codec_ops *codec_ops = sess->fmt_out->codec_ops; + + if (core->platform->revision == VDEC_REVISION_G12A || + core->platform->revision == VDEC_REVISION_SM1) { + clk_set_rate(core->vdec_hevcf_clk, 666666666); + ret = clk_prepare_enable(core->vdec_hevcf_clk); + if (ret) + return ret; + } + + clk_set_rate(core->vdec_hevc_clk, 666666666); + ret = clk_prepare_enable(core->vdec_hevc_clk); + if (ret) + return ret; + + if (core->platform->revision == VDEC_REVISION_SM1) + regmap_update_bits(core->regmap_ao, AO_RTI_GEN_PWR_SLEEP0, + GEN_PWR_VDEC_HEVC_SM1, 0); + else + regmap_update_bits(core->regmap_ao, AO_RTI_GEN_PWR_SLEEP0, + GEN_PWR_VDEC_HEVC, 0); + udelay(10); + + /* Reset VDEC_HEVC*/ + amvdec_write_dos(core, DOS_SW_RESET3, 0xffffffff); + amvdec_write_dos(core, DOS_SW_RESET3, 0x00000000); + + amvdec_write_dos(core, DOS_GCLK_EN3, 0xffffffff); + + /* VDEC_HEVC Memories */ + amvdec_write_dos(core, DOS_MEM_PD_HEVC, 0x00000000); + + /* Remove VDEC_HEVC Isolation */ + if (core->platform->revision == VDEC_REVISION_SM1) + regmap_update_bits(core->regmap_ao, AO_RTI_GEN_PWR_ISO0, + GEN_PWR_VDEC_HEVC_SM1, 0); + else + regmap_update_bits(core->regmap_ao, AO_RTI_GEN_PWR_ISO0, + 0xc00, 0); + + amvdec_write_dos(core, DOS_SW_RESET3, 0xffffffff); + amvdec_write_dos(core, DOS_SW_RESET3, 0x00000000); + + vdec_hevc_stbuf_init(sess); + + ret = vdec_hevc_load_firmware(sess, sess->fmt_out->firmware_path); + if (ret) + goto stop; + + ret = codec_ops->start(sess); + if (ret) + goto stop; + + amvdec_write_dos(core, DOS_SW_RESET3, BIT(12) | BIT(11)); + amvdec_write_dos(core, DOS_SW_RESET3, 0); + amvdec_read_dos(core, DOS_SW_RESET3); + + amvdec_write_dos(core, HEVC_MPSR, 1); + /* Let the firmware settle */ + udelay(10); + + return 0; + +stop: + vdec_hevc_stop(sess); + return ret; +} + +struct amvdec_ops vdec_hevc_ops = { + .start = vdec_hevc_start, + .stop = vdec_hevc_stop, + .conf_esparser = vdec_hevc_conf_esparser, + .vififo_level = vdec_hevc_vififo_level, +}; diff --git a/drivers/staging/media/meson/vdec/vdec_hevc.h b/drivers/staging/media/meson/vdec/vdec_hevc.h new file mode 100644 index 000000000000..cd576a73a966 --- /dev/null +++ b/drivers/staging/media/meson/vdec/vdec_hevc.h @@ -0,0 +1,13 @@ +/* SPDX-License-Identifier: GPL-2.0+ */ +/* + * Copyright (C) 2018 Maxime Jourdan + */ + +#ifndef __MESON_VDEC_VDEC_HEVC_H_ +#define __MESON_VDEC_VDEC_HEVC_H_ + +#include "vdec.h" + +extern struct amvdec_ops vdec_hevc_ops; + +#endif From patchwork Thu Dec 5 09:26:39 2019 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Neil Armstrong X-Patchwork-Id: 11274409 Return-Path: Received: from mail.kernel.org (pdx-korg-mail-1.web.codeaurora.org [172.30.200.123]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id 53D5714BD for ; Thu, 5 Dec 2019 09:26:46 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 2779B24653 for ; Thu, 5 Dec 2019 09:26:46 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=pass (2048-bit key) header.d=baylibre-com.20150623.gappssmtp.com header.i=@baylibre-com.20150623.gappssmtp.com header.b="ReLcbyzB" Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1728892AbfLEJ0p (ORCPT ); Thu, 5 Dec 2019 04:26:45 -0500 Received: from mail-wr1-f65.google.com ([209.85.221.65]:41097 "EHLO mail-wr1-f65.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1726096AbfLEJ0p (ORCPT ); Thu, 5 Dec 2019 04:26:45 -0500 Received: by mail-wr1-f65.google.com with SMTP id c9so2572775wrw.8 for ; Thu, 05 Dec 2019 01:26:43 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=baylibre-com.20150623.gappssmtp.com; s=20150623; h=from:to:cc:subject:date:message-id:in-reply-to:references :mime-version:content-transfer-encoding; bh=NGrwUpvZsufFSsKAC9GKZZNPeviBTCu0uYbSSoXF3hw=; b=ReLcbyzBiUCQ/ScDPDv0aYOdptMXiafQXXYxY2s/E96YdEm75sl6PbYm+YGtuJEOEc HOqsBLOQtdMkbjKJOfloDrukTINEo37QKQKOoA0AIP9jowIdNp1g4kAJEobVHIiJ5/41 cIVHRakpg8oq1xA0Zf72d0WPFraR+CMxT0U/e8AVIho/WhWDps+c91cJnT08ciAsVJan MTsZuV9NBt+hE87Pq/gFUGnXPVI6Nl1Zg5gVhfoDsUxe5WVsoqPB+uGeKsp70XfbLhlC N9JsimtV0IhwKuFGzWosrRaWsDbNgpwu7dscGN4krIpf7paAxjJBPmIr3O5GV7Njs+22 fSLw== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to :references:mime-version:content-transfer-encoding; bh=NGrwUpvZsufFSsKAC9GKZZNPeviBTCu0uYbSSoXF3hw=; b=i1ntLi3d/VdD4+76BHQWaKwPVmU9njQZ59s13osJ0yLI2yGaD8nY1BK6irgrbtPf7p Txy+27Rk7SYrFZBmLVY2rVWjATLZ0Es2Eynop4NhV6/r+xnTzgzgvI1v1LGlyUtIG4lc LivFTSvGV9SmhRtNVjG/9cUWxW0g6hkvHVP6PxJyV8/t3nyDdoa/vCNxnNCmGs5W+iVK rQIqK5xDpemJyp+c+IMJirRwZaSOjGMUmrk2V9RCpQQA3uWfMIoIj187HHqzixYHx6ma zcKtHaLr9sMCX4hhvC2puQhnoqb9HEBV4g7P7JUSP6gVjFT8xqghgEdiVmvusnCP1TO6 mPeg== X-Gm-Message-State: APjAAAXoxgQHpq9kP1zpQIsCmMv3tIuV2ZvNL/9yPfhDLcvs2v0zfJKQ 2FuP0mSZNiajK0K+AtPahV35BQ== X-Google-Smtp-Source: APXvYqzmj2qKKKkQa1ZZ8hRrZhOqTNNsZkVqFjt3ttGrZWbut+nhaZvlckkRDHcWX0lhvEolL77+oA== X-Received: by 2002:adf:fcc4:: with SMTP id f4mr322411wrs.247.1575538002082; Thu, 05 Dec 2019 01:26:42 -0800 (PST) Received: from bender.baylibre.local (wal59-h01-176-150-251-154.dsl.sta.abo.bbox.fr. [176.150.251.154]) by smtp.gmail.com with ESMTPSA id v4sm1264780wml.2.2019.12.05.01.26.41 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Thu, 05 Dec 2019 01:26:41 -0800 (PST) From: Neil Armstrong To: mchehab@kernel.org, hans.verkuil@cisco.com Cc: Maxime Jourdan , linux-media@vger.kernel.org, linux-amlogic@lists.infradead.org, linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org, Neil Armstrong Subject: [PATCH 4/5] media: meson: vdec: add VP9 input support Date: Thu, 5 Dec 2019 10:26:39 +0100 Message-Id: <20191205092639.26330-1-narmstrong@baylibre.com> X-Mailer: git-send-email 2.22.0 In-Reply-To: <20191205092454.26075-1-narmstrong@baylibre.com> References: <20191205092454.26075-1-narmstrong@baylibre.com> MIME-Version: 1.0 Sender: linux-media-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-media@vger.kernel.org From: Maxime Jourdan Amlogic VP9 decoder requires an additional 16-byte payload before every frame header. Signed-off-by: Maxime Jourdan Signed-off-by: Neil Armstrong --- drivers/staging/media/meson/vdec/esparser.c | 142 +++++++++++++++++++- 1 file changed, 138 insertions(+), 4 deletions(-) diff --git a/drivers/staging/media/meson/vdec/esparser.c b/drivers/staging/media/meson/vdec/esparser.c index adc5c1e81a4c..aeb68f6c732a 100644 --- a/drivers/staging/media/meson/vdec/esparser.c +++ b/drivers/staging/media/meson/vdec/esparser.c @@ -52,6 +52,7 @@ #define PARSER_VIDEO_HOLE 0x90 #define SEARCH_PATTERN_LEN 512 +#define VP9_HEADER_SIZE 16 static DECLARE_WAIT_QUEUE_HEAD(wq); static int search_done; @@ -74,14 +75,121 @@ static irqreturn_t esparser_isr(int irq, void *dev) return IRQ_HANDLED; } +/** + * VP9 frame headers need to be appended by a 16-byte long + * Amlogic custom header + */ +static int vp9_update_header(struct amvdec_core *core, struct vb2_buffer *buf) +{ + u8 *dp; + u8 marker; + int dsize; + int num_frames, cur_frame; + int cur_mag, mag, mag_ptr; + int frame_size[8], tot_frame_size[8]; + int total_datasize = 0; + int new_frame_size; + unsigned char *old_header = NULL; + + dp = (uint8_t *)vb2_plane_vaddr(buf, 0); + dsize = vb2_get_plane_payload(buf, 0); + + if (dsize == vb2_plane_size(buf, 0)) { + dev_warn(core->dev, "%s: unable to update header\n", __func__); + return 0; + } + + marker = dp[dsize - 1]; + if ((marker & 0xe0) == 0xc0) { + num_frames = (marker & 0x7) + 1; + mag = ((marker >> 3) & 0x3) + 1; + mag_ptr = dsize - mag * num_frames - 2; + if (dp[mag_ptr] != marker) + return 0; + + mag_ptr++; + for (cur_frame = 0; cur_frame < num_frames; cur_frame++) { + frame_size[cur_frame] = 0; + for (cur_mag = 0; cur_mag < mag; cur_mag++) { + frame_size[cur_frame] |= + (dp[mag_ptr] << (cur_mag * 8)); + mag_ptr++; + } + if (cur_frame == 0) + tot_frame_size[cur_frame] = + frame_size[cur_frame]; + else + tot_frame_size[cur_frame] = + tot_frame_size[cur_frame - 1] + + frame_size[cur_frame]; + total_datasize += frame_size[cur_frame]; + } + } else { + num_frames = 1; + frame_size[0] = dsize; + tot_frame_size[0] = dsize; + total_datasize = dsize; + } + + new_frame_size = total_datasize + num_frames * VP9_HEADER_SIZE; + + if (new_frame_size >= vb2_plane_size(buf, 0)) { + dev_warn(core->dev, "%s: unable to update header\n", __func__); + return 0; + } + + for (cur_frame = num_frames - 1; cur_frame >= 0; cur_frame--) { + int framesize = frame_size[cur_frame]; + int framesize_header = framesize + 4; + int oldframeoff = tot_frame_size[cur_frame] - framesize; + int outheaderoff = oldframeoff + cur_frame * VP9_HEADER_SIZE; + u8 *fdata = dp + outheaderoff; + u8 *old_framedata = dp + oldframeoff; + + memmove(fdata + VP9_HEADER_SIZE, old_framedata, framesize); + + fdata[0] = (framesize_header >> 24) & 0xff; + fdata[1] = (framesize_header >> 16) & 0xff; + fdata[2] = (framesize_header >> 8) & 0xff; + fdata[3] = (framesize_header >> 0) & 0xff; + fdata[4] = ((framesize_header >> 24) & 0xff) ^ 0xff; + fdata[5] = ((framesize_header >> 16) & 0xff) ^ 0xff; + fdata[6] = ((framesize_header >> 8) & 0xff) ^ 0xff; + fdata[7] = ((framesize_header >> 0) & 0xff) ^ 0xff; + fdata[8] = 0; + fdata[9] = 0; + fdata[10] = 0; + fdata[11] = 1; + fdata[12] = 'A'; + fdata[13] = 'M'; + fdata[14] = 'L'; + fdata[15] = 'V'; + + if (!old_header) { + /* nothing */ + } else if (old_header > fdata + 16 + framesize) { + dev_dbg(core->dev, "%s: data has gaps, setting to 0\n", + __func__); + memset(fdata + 16 + framesize, 0, + (old_header - fdata + 16 + framesize)); + } else if (old_header < fdata + 16 + framesize) { + dev_err(core->dev, "%s: data overwritten\n", __func__); + } + old_header = fdata; + } + + return new_frame_size; +} + /* Pad the packet to at least 4KiB bytes otherwise the VDEC unit won't trigger * ISRs. * Also append a start code 000001ff at the end to trigger * the ESPARSER interrupt. */ -static u32 esparser_pad_start_code(struct amvdec_core *core, struct vb2_buffer *vb) +static u32 esparser_pad_start_code(struct amvdec_core *core, + struct vb2_buffer *vb, + u32 payload_size) { - u32 payload_size = vb2_get_plane_payload(vb, 0); u32 pad_size = 0; u8 *vaddr = vb2_plane_vaddr(vb, 0); @@ -186,13 +294,27 @@ esparser_queue(struct amvdec_session *sess, struct vb2_v4l2_buffer *vbuf) int ret; struct vb2_buffer *vb = &vbuf->vb2_buf; struct amvdec_core *core = sess->core; + struct amvdec_codec_ops *codec_ops = sess->fmt_out->codec_ops; u32 payload_size = vb2_get_plane_payload(vb, 0); dma_addr_t phy = vb2_dma_contig_plane_dma_addr(vb, 0); + u32 num_dst_bufs = 0; u32 offset; u32 pad_size; - if (esparser_vififo_get_free_space(sess) < payload_size) + if (sess->fmt_out->pixfmt == V4L2_PIX_FMT_VP9) { + if (codec_ops->num_pending_bufs) + num_dst_bufs = codec_ops->num_pending_bufs(sess); + + num_dst_bufs += v4l2_m2m_num_dst_bufs_ready(sess->m2m_ctx); + if (sess->fmt_out->pixfmt == V4L2_PIX_FMT_VP9) + num_dst_bufs -= 2; + + if (esparser_vififo_get_free_space(sess) < payload_size || + atomic_read(&sess->esparser_queued_bufs) >= num_dst_bufs) + return -EAGAIN; + } else if (esparser_vififo_get_free_space(sess) < payload_size) { return -EAGAIN; + } v4l2_m2m_src_buf_remove_by_buf(sess->m2m_ctx, vbuf); @@ -206,7 +328,19 @@ esparser_queue(struct amvdec_session *sess, struct vb2_v4l2_buffer *vbuf) vbuf->field = V4L2_FIELD_NONE; vbuf->sequence = sess->sequence_out++; - pad_size = esparser_pad_start_code(core, vb); + if (sess->fmt_out->pixfmt == V4L2_PIX_FMT_VP9) { + payload_size = vp9_update_header(core, vb); + + /* If unable to alter buffer to add headers */ + if (payload_size == 0) { + amvdec_remove_ts(sess, vb->timestamp); + v4l2_m2m_buf_done(vbuf, VB2_BUF_STATE_ERROR); + + return 0; + } + } + + pad_size = esparser_pad_start_code(core, vb, payload_size); ret = esparser_write_data(core, phy, payload_size + pad_size); if (ret <= 0) { From patchwork Thu Dec 5 09:26:49 2019 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Neil Armstrong X-Patchwork-Id: 11274415 Return-Path: Received: from mail.kernel.org (pdx-korg-mail-1.web.codeaurora.org [172.30.200.123]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id 136D513A4 for ; Thu, 5 Dec 2019 09:27:00 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id BE7C524655 for ; Thu, 5 Dec 2019 09:26:59 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=pass (2048-bit key) header.d=baylibre-com.20150623.gappssmtp.com header.i=@baylibre-com.20150623.gappssmtp.com header.b="R5H/vrKw" Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1729091AbfLEJ06 (ORCPT ); Thu, 5 Dec 2019 04:26:58 -0500 Received: from mail-wr1-f65.google.com ([209.85.221.65]:35655 "EHLO mail-wr1-f65.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1726239AbfLEJ05 (ORCPT ); Thu, 5 Dec 2019 04:26:57 -0500 Received: by mail-wr1-f65.google.com with SMTP id g17so2619442wro.2 for ; Thu, 05 Dec 2019 01:26:52 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=baylibre-com.20150623.gappssmtp.com; s=20150623; h=from:to:cc:subject:date:message-id:in-reply-to:references :mime-version:content-transfer-encoding; bh=rhXs7fJVG1WdVXuA//i2y/njlm/IzJuwom6kVvOJzlk=; b=R5H/vrKwdm7ABLxpXQPEth98cH897nlSTjFMaU/hjp3QjuNvpXqEK7hAWc+tS0y0Zm pahkHXx3spVN0tIEWpY6izSnwSJN9r1tgnKq+sF0FoXRJ76mQpyqGHWgRgxshwiW9hAM VKCfJRgG8nXFydYNx09f3OTl+n2ZwPKlBp/uZ8AmwT+qKRR9R94jFbMkkG+sVk4gWRPP moQA7O1cB0b+B4WWIBvjnBK5rxKCe7NBB5TWK+6LGnczCx1DBZA0EgAcR3VIrK+GkC8h N9yf41WsbTykEG2AiCnv29mSFKkAUvK1/xOhSeLRCGvzE8+2fXOBdyz1Q6mFyVRiCG9E NMmA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to :references:mime-version:content-transfer-encoding; bh=rhXs7fJVG1WdVXuA//i2y/njlm/IzJuwom6kVvOJzlk=; b=DsJoiHdmZTvMpDpscMcx5xiKQ9GwGELD49ivH9xEmPCVPrpfq07DwCUPJXGOLO5k74 VuCohzrBmKu1RTl3/FLVymesuZ5Fk3IDJKY7UCYqa00G01AGDYZpRSI+28Hyp74XQbT1 +uYvqmOF74P9NKwjlW68APKM1cDgTV8c96Gnbr4YEr5pbb6NvVc7JXzfgKFBBt2HI5Rs HOH3AGKJy03CPmRzUV923JmW5JojioXnNWosCftwcxAkamQ2k/S2x3xvFWb6fhYX5pKd WOPAIbdhp7hO0E6qt/1kKTQYXib0ONGVr+6645BOciazu+3YD75M2Ztv88t8w6NP8DQ9 rv0Q== X-Gm-Message-State: APjAAAXnLumJk9vochvx5cC5ifjGL58xlFttKjtAlXgS8SymqpszQH61 +RpHzemproOSPR1i27iu8++WzA== X-Google-Smtp-Source: APXvYqwElrscBOj3K4BAX3LdUHDUTUOuibpcG73joa/ADMEHSUBM9AMXipAsaPlpwfMIEZK4dDnHqA== X-Received: by 2002:adf:bc87:: with SMTP id g7mr9265325wrh.121.1575538010989; Thu, 05 Dec 2019 01:26:50 -0800 (PST) Received: from bender.baylibre.local (wal59-h01-176-150-251-154.dsl.sta.abo.bbox.fr. [176.150.251.154]) by smtp.gmail.com with ESMTPSA id u14sm11252183wrm.51.2019.12.05.01.26.50 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Thu, 05 Dec 2019 01:26:50 -0800 (PST) From: Neil Armstrong To: mchehab@kernel.org, hans.verkuil@cisco.com Cc: Maxime Jourdan , linux-media@vger.kernel.org, linux-amlogic@lists.infradead.org, linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org, Neil Armstrong Subject: [PATCH 5/5] media: meson: vdec: add VP9 decoder support Date: Thu, 5 Dec 2019 10:26:49 +0100 Message-Id: <20191205092649.26382-1-narmstrong@baylibre.com> X-Mailer: git-send-email 2.22.0 In-Reply-To: <20191205092454.26075-1-narmstrong@baylibre.com> References: <20191205092454.26075-1-narmstrong@baylibre.com> MIME-Version: 1.0 Sender: linux-media-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-media@vger.kernel.org From: Maxime Jourdan This adds VP9 decoding for the Amlogic GXL, G12A & SM1 SoCs, using the commong "HEVC" HW decoder. For G12A & SM1, it uses the IOMMU support from the firmware. For 10bit decoding, the firmware can only decode in the proprietary Amlogic Framebuffer Compression format, but can output in 8bit NV12 buffer while writing the decoded frame. Signed-off-by: Maxime Jourdan Signed-off-by: Neil Armstrong --- drivers/staging/media/meson/vdec/Makefile | 2 +- drivers/staging/media/meson/vdec/codec_vp9.c | 1192 +++++++++++++++++ drivers/staging/media/meson/vdec/codec_vp9.h | 13 + drivers/staging/media/meson/vdec/hevc_regs.h | 7 + .../staging/media/meson/vdec/vdec_platform.c | 38 + 5 files changed, 1251 insertions(+), 1 deletion(-) create mode 100644 drivers/staging/media/meson/vdec/codec_vp9.c create mode 100644 drivers/staging/media/meson/vdec/codec_vp9.h diff --git a/drivers/staging/media/meson/vdec/Makefile b/drivers/staging/media/meson/vdec/Makefile index f55b6e625034..6e726af84ac9 100644 --- a/drivers/staging/media/meson/vdec/Makefile +++ b/drivers/staging/media/meson/vdec/Makefile @@ -3,6 +3,6 @@ meson-vdec-objs = esparser.o vdec.o vdec_helpers.o vdec_platform.o meson-vdec-objs += vdec_1.o vdec_hevc.o -meson-vdec-objs += codec_mpeg12.o codec_h264.o codec_hevc_common.o +meson-vdec-objs += codec_mpeg12.o codec_h264.o codec_hevc_common.o codec_vp9.o obj-$(CONFIG_VIDEO_MESON_VDEC) += meson-vdec.o diff --git a/drivers/staging/media/meson/vdec/codec_vp9.c b/drivers/staging/media/meson/vdec/codec_vp9.c new file mode 100644 index 000000000000..3f17188d9675 --- /dev/null +++ b/drivers/staging/media/meson/vdec/codec_vp9.c @@ -0,0 +1,1192 @@ +// SPDX-License-Identifier: GPL-2.0+ +/* + * Copyright (C) 2018 Maxime Jourdan + * Copyright (C) 2015 Amlogic, Inc. All rights reserved. + */ + +#include +#include + +#include "dos_regs.h" +#include "hevc_regs.h" +#include "vdec_helpers.h" +#include "codec_hevc_common.h" + +/* HEVC reg mapping */ +#define VP9_DEC_STATUS_REG HEVC_ASSIST_SCRATCH_0 + #define VP9_10B_DECODE_SLICE 5 + #define VP9_HEAD_PARSER_DONE 0xf0 +#define VP9_RPM_BUFFER HEVC_ASSIST_SCRATCH_1 +#define VP9_SHORT_TERM_RPS HEVC_ASSIST_SCRATCH_2 +#define VP9_ADAPT_PROB_REG HEVC_ASSIST_SCRATCH_3 +#define VP9_MMU_MAP_BUFFER HEVC_ASSIST_SCRATCH_4 +#define VP9_PPS_BUFFER HEVC_ASSIST_SCRATCH_5 +#define VP9_SAO_UP HEVC_ASSIST_SCRATCH_6 +#define VP9_STREAM_SWAP_BUFFER HEVC_ASSIST_SCRATCH_7 +#define VP9_STREAM_SWAP_BUFFER2 HEVC_ASSIST_SCRATCH_8 +#define VP9_PROB_SWAP_BUFFER HEVC_ASSIST_SCRATCH_9 +#define VP9_COUNT_SWAP_BUFFER HEVC_ASSIST_SCRATCH_A +#define VP9_SEG_MAP_BUFFER HEVC_ASSIST_SCRATCH_B +#define VP9_SCALELUT HEVC_ASSIST_SCRATCH_D +#define VP9_WAIT_FLAG HEVC_ASSIST_SCRATCH_E +#define LMEM_DUMP_ADR HEVC_ASSIST_SCRATCH_F +#define NAL_SEARCH_CTL HEVC_ASSIST_SCRATCH_I +#define VP9_DECODE_MODE HEVC_ASSIST_SCRATCH_J + #define DECODE_MODE_SINGLE 0 +#define DECODE_STOP_POS HEVC_ASSIST_SCRATCH_K +#define HEVC_DECODE_COUNT HEVC_ASSIST_SCRATCH_M +#define HEVC_DECODE_SIZE HEVC_ASSIST_SCRATCH_N + +/* VP9 Constants */ +#define LCU_SIZE 64 +#define MAX_REF_PIC_NUM 24 +#define REFS_PER_FRAME 3 +#define REF_FRAMES 8 +#define MV_MEM_UNIT 0x240 + +enum FRAME_TYPE { + KEY_FRAME = 0, + INTER_FRAME = 1, + FRAME_TYPES, +}; + +/* VP9 Workspace layout */ +#define MPRED_MV_BUF_SIZE 0x120000 + +#define IPP_SIZE 0x4000 +#define SAO_ABV_SIZE 0x30000 +#define SAO_VB_SIZE 0x30000 +#define SH_TM_RPS_SIZE 0x800 +#define VPS_SIZE 0x800 +#define SPS_SIZE 0x800 +#define PPS_SIZE 0x2000 +#define SAO_UP_SIZE 0x2800 +#define SWAP_BUF_SIZE 0x800 +#define SWAP_BUF2_SIZE 0x800 +#define SCALELUT_SIZE 0x8000 +#define DBLK_PARA_SIZE 0x80000 +#define DBLK_DATA_SIZE 0x80000 +#define SEG_MAP_SIZE 0xd800 +#define PROB_SIZE 0x5000 +#define COUNT_SIZE 0x3000 +#define MMU_VBH_SIZE 0x5000 +#define MPRED_ABV_SIZE 0x10000 +#define MPRED_MV_SIZE (MPRED_MV_BUF_SIZE * MAX_REF_PIC_NUM) +#define RPM_BUF_SIZE 0x100 +#define LMEM_SIZE 0x800 + +#define IPP_OFFSET 0x00 +#define SAO_ABV_OFFSET (IPP_OFFSET + IPP_SIZE) +#define SAO_VB_OFFSET (SAO_ABV_OFFSET + SAO_ABV_SIZE) +#define SH_TM_RPS_OFFSET (SAO_VB_OFFSET + SAO_VB_SIZE) +#define VPS_OFFSET (SH_TM_RPS_OFFSET + SH_TM_RPS_SIZE) +#define SPS_OFFSET (VPS_OFFSET + VPS_SIZE) +#define PPS_OFFSET (SPS_OFFSET + SPS_SIZE) +#define SAO_UP_OFFSET (PPS_OFFSET + PPS_SIZE) +#define SWAP_BUF_OFFSET (SAO_UP_OFFSET + SAO_UP_SIZE) +#define SWAP_BUF2_OFFSET (SWAP_BUF_OFFSET + SWAP_BUF_SIZE) +#define SCALELUT_OFFSET (SWAP_BUF2_OFFSET + SWAP_BUF2_SIZE) +#define DBLK_PARA_OFFSET (SCALELUT_OFFSET + SCALELUT_SIZE) +#define DBLK_DATA_OFFSET (DBLK_PARA_OFFSET + DBLK_PARA_SIZE) +#define SEG_MAP_OFFSET (DBLK_DATA_OFFSET + DBLK_DATA_SIZE) +#define PROB_OFFSET (SEG_MAP_OFFSET + SEG_MAP_SIZE) +#define COUNT_OFFSET (PROB_OFFSET + PROB_SIZE) +#define MMU_VBH_OFFSET (COUNT_OFFSET + COUNT_SIZE) +#define MPRED_ABV_OFFSET (MMU_VBH_OFFSET + MMU_VBH_SIZE) +#define MPRED_MV_OFFSET (MPRED_ABV_OFFSET + MPRED_ABV_SIZE) +#define RPM_OFFSET (MPRED_MV_OFFSET + MPRED_MV_SIZE) +#define LMEM_OFFSET (RPM_OFFSET + RPM_BUF_SIZE) + +#define SIZE_WORKSPACE ALIGN(LMEM_OFFSET + LMEM_SIZE, 64 * SZ_1K) + +#define NONE -1 +#define INTRA_FRAME 0 +#define LAST_FRAME 1 +#define GOLDEN_FRAME 2 +#define ALTREF_FRAME 3 +#define MAX_REF_FRAMES 4 + +/* + * Defines, declarations, sub-functions for vp9 de-block loop + filter Thr/Lvl table update + * - struct segmentation is for loop filter only (removed something) + * - function "vp9_loop_filter_init" and "vp9_loop_filter_frame_init" will + be instantiated in C_Entry + * - vp9_loop_filter_init run once before decoding start + * - vp9_loop_filter_frame_init run before every frame decoding start + * - set video format to VP9 is in vp9_loop_filter_init + */ +#define MAX_LOOP_FILTER 63 +#define MAX_REF_LF_DELTAS 4 +#define MAX_MODE_LF_DELTAS 2 +#define SEGMENT_DELTADATA 0 +#define SEGMENT_ABSDATA 1 +#define MAX_SEGMENTS 8 + +union rpm_param { + struct { + u16 data[RPM_BUF_SIZE]; + } l; + struct { + u16 profile; + u16 show_existing_frame; + u16 frame_to_show_idx; + u16 frame_type; /*1 bit*/ + u16 show_frame; /*1 bit*/ + u16 error_resilient_mode; /*1 bit*/ + u16 intra_only; /*1 bit*/ + u16 display_size_present; /*1 bit*/ + u16 reset_frame_context; + u16 refresh_frame_flags; + u16 width; + u16 height; + u16 display_width; + u16 display_height; + u16 ref_info; + u16 same_frame_size; + u16 mode_ref_delta_enabled; + u16 ref_deltas[4]; + u16 mode_deltas[2]; + u16 filter_level; + u16 sharpness_level; + u16 bit_depth; + u16 seg_quant_info[8]; + u16 seg_enabled; + u16 seg_abs_delta; + /* bit 15: feature enabled; bit 8, sign; bit[5:0], data */ + u16 seg_lf_info[8]; + } p; +}; + +enum SEG_LVL_FEATURES { + SEG_LVL_ALT_Q = 0, /* Use alternate Quantizer */ + SEG_LVL_ALT_LF = 1, /* Use alternate loop filter value */ + SEG_LVL_REF_FRAME = 2, /* Optional Segment reference frame */ + SEG_LVL_SKIP = 3, /* Optional Segment (0,0) + skip mode */ + SEG_LVL_MAX = 4 /* Number of features supported */ +}; + +struct segmentation { + u8 enabled; + u8 update_map; + u8 update_data; + u8 abs_delta; + u8 temporal_update; + s16 feature_data[MAX_SEGMENTS][SEG_LVL_MAX]; + unsigned int feature_mask[MAX_SEGMENTS]; +}; + +struct loop_filter_thresh { + u8 mblim; + u8 lim; + u8 hev_thr; +}; + +struct loop_filter_info_n { + struct loop_filter_thresh lfthr[MAX_LOOP_FILTER + 1]; + u8 lvl[MAX_SEGMENTS][MAX_REF_FRAMES][MAX_MODE_LF_DELTAS]; +}; + +struct loopfilter { + int filter_level; + + int sharpness_level; + int last_sharpness_level; + + u8 mode_ref_delta_enabled; + u8 mode_ref_delta_update; + + /*0 = Intra, Last, GF, ARF*/ + signed char ref_deltas[MAX_REF_LF_DELTAS]; + signed char last_ref_deltas[MAX_REF_LF_DELTAS]; + + /*0 = ZERO_MV, MV*/ + signed char mode_deltas[MAX_MODE_LF_DELTAS]; + signed char last_mode_deltas[MAX_MODE_LF_DELTAS]; +}; + +struct vp9_frame { + struct list_head list; + struct vb2_v4l2_buffer *vbuf; + int index; + int intra_only; + int show; + int type; + int done; +}; + +struct codec_vp9 { + /* VP9 context lock */ + struct mutex lock; + + /* Common part with the HEVC decoder */ + struct codec_hevc_common common; + + /* Buffer for the VP9 Workspace */ + void *workspace_vaddr; + dma_addr_t workspace_paddr; + + /* Contains many information parsed from the bitstream */ + union rpm_param rpm_param; + + /* Whether we detected the bitstream as 10-bit */ + int is_10bit; + + /* Coded resolution reported by the hardware */ + u32 width, height; + + /* All ref frames used by the HW at a given time */ + struct list_head ref_frames_list; + u32 frames_num; + + /* In case of downsampling (decoding with FBC but outputting in NV12M), + * we need to allocate additional buffers for FBC. + */ + void *fbc_buffer_vaddr[MAX_REF_PIC_NUM]; + dma_addr_t fbc_buffer_paddr[MAX_REF_PIC_NUM]; + + int ref_frame_map[REF_FRAMES]; + int next_ref_frame_map[REF_FRAMES]; + struct vp9_frame *frame_refs[REFS_PER_FRAME]; + + u32 lcu_total; + + /* loop filter */ + int default_filt_lvl; + struct loop_filter_info_n lfi; + struct loopfilter lf; + struct segmentation seg_4lf; + + struct vp9_frame *cur_frame; + struct vp9_frame *prev_frame; +}; + +static int vp9_clamp(int value, int low, int high) +{ + return value < low ? low : (value > high ? high : value); +} + +static int segfeature_active(struct segmentation *seg, int segment_id, + enum SEG_LVL_FEATURES feature_id) +{ + return seg->enabled && + (seg->feature_mask[segment_id] & (1 << feature_id)); +} + +static int get_segdata(struct segmentation *seg, int segment_id, + enum SEG_LVL_FEATURES feature_id) +{ + return seg->feature_data[segment_id][feature_id]; +} + +static void vp9_update_sharpness(struct loop_filter_info_n *lfi, + int sharpness_lvl) +{ + int lvl; + + /* For each possible value for the loop filter fill out limits*/ + for (lvl = 0; lvl <= MAX_LOOP_FILTER; lvl++) { + /* Set loop filter parameters that control sharpness.*/ + int block_inside_limit = lvl >> ((sharpness_lvl > 0) + + (sharpness_lvl > 4)); + + if (sharpness_lvl > 0) { + if (block_inside_limit > (9 - sharpness_lvl)) + block_inside_limit = (9 - sharpness_lvl); + } + + if (block_inside_limit < 1) + block_inside_limit = 1; + + lfi->lfthr[lvl].lim = (u8)block_inside_limit; + lfi->lfthr[lvl].mblim = (u8)(2 * (lvl + 2) + + block_inside_limit); + } +} + +/* Instantiate this function once when decode is started */ +static void +vp9_loop_filter_init(struct amvdec_core *core, struct codec_vp9 *vp9) +{ + struct loop_filter_info_n *lfi = &vp9->lfi; + struct loopfilter *lf = &vp9->lf; + struct segmentation *seg_4lf = &vp9->seg_4lf; + int i; + + memset(lfi, 0, sizeof(struct loop_filter_info_n)); + memset(lf, 0, sizeof(struct loopfilter)); + memset(seg_4lf, 0, sizeof(struct segmentation)); + lf->sharpness_level = 0; + vp9_update_sharpness(lfi, lf->sharpness_level); + lf->last_sharpness_level = lf->sharpness_level; + + for (i = 0; i < 32; i++) { + unsigned int thr; + + thr = ((lfi->lfthr[i * 2 + 1].lim & 0x3f) << 8) | + (lfi->lfthr[i * 2 + 1].mblim & 0xff); + thr = (thr << 16) | ((lfi->lfthr[i * 2].lim & 0x3f) << 8) | + (lfi->lfthr[i * 2].mblim & 0xff); + + amvdec_write_dos(core, HEVC_DBLK_CFG9, thr); + } + + if (core->platform->revision >= VDEC_REVISION_G12A) + /* VP9 video format */ + amvdec_write_dos(core, HEVC_DBLK_CFGB, (0x54 << 8) | BIT(0)); + else if (core->platform->revision >= VDEC_REVISION_SM1) + amvdec_write_dos(core, HEVC_DBLK_CFGB, + (0x3 << 14) | /* dw fifo thres r and b */ + (0x3 << 12) | /* dw fifo thres r or b */ + (0x3 << 10) | /* dw fifo thres not r/b */ + BIT(0)); /* VP9 video format */ + else + amvdec_write_dos(core, HEVC_DBLK_CFGB, 0x40400001); +} + +static void +vp9_loop_filter_frame_init(struct amvdec_core *core, struct segmentation *seg, + struct loop_filter_info_n *lfi, + struct loopfilter *lf, int default_filt_lvl) +{ + int i; + int seg_id; + + /* + * n_shift is the multiplier for lf_deltas + * the multiplier is: + * - 1 for when filter_lvl is between 0 and 31 + * - 2 when filter_lvl is between 32 and 63 + */ + const int scale = 1 << (default_filt_lvl >> 5); + + /* update limits if sharpness has changed */ + if (lf->last_sharpness_level != lf->sharpness_level) { + vp9_update_sharpness(lfi, lf->sharpness_level); + lf->last_sharpness_level = lf->sharpness_level; + + /* Write to register */ + for (i = 0; i < 32; i++) { + unsigned int thr; + + thr = ((lfi->lfthr[i * 2 + 1].lim & 0x3f) << 8) | + (lfi->lfthr[i * 2 + 1].mblim & 0xff); + thr = (thr << 16) | + ((lfi->lfthr[i * 2].lim & 0x3f) << 8) | + (lfi->lfthr[i * 2].mblim & 0xff); + + amvdec_write_dos(core, HEVC_DBLK_CFG9, thr); + } + } + + for (seg_id = 0; seg_id < MAX_SEGMENTS; seg_id++) { + int lvl_seg = default_filt_lvl; + + if (segfeature_active(seg, seg_id, SEG_LVL_ALT_LF)) { + const int data = get_segdata(seg, seg_id, + SEG_LVL_ALT_LF); + lvl_seg = vp9_clamp(seg->abs_delta == SEGMENT_ABSDATA ? + data : default_filt_lvl + data, + 0, MAX_LOOP_FILTER); + } + + if (!lf->mode_ref_delta_enabled) { + /* + * We could get rid of this if we assume that deltas + * are set to zero when not in use. + * encoder always uses deltas + */ + memset(lfi->lvl[seg_id], lvl_seg, + sizeof(lfi->lvl[seg_id])); + } else { + int ref, mode; + const int intra_lvl = + lvl_seg + lf->ref_deltas[INTRA_FRAME] * scale; + lfi->lvl[seg_id][INTRA_FRAME][0] = + vp9_clamp(intra_lvl, 0, MAX_LOOP_FILTER); + + for (ref = LAST_FRAME; ref < MAX_REF_FRAMES; ++ref) { + for (mode = 0; mode < MAX_MODE_LF_DELTAS; + ++mode) { + const int inter_lvl = + lvl_seg + + lf->ref_deltas[ref] * scale + + lf->mode_deltas[mode] * scale; + lfi->lvl[seg_id][ref][mode] = + vp9_clamp(inter_lvl, 0, + MAX_LOOP_FILTER); + } + } + } + } + + for (i = 0; i < 16; i++) { + unsigned int level; + + level = ((lfi->lvl[i >> 1][3][i & 1] & 0x3f) << 24) | + ((lfi->lvl[i >> 1][2][i & 1] & 0x3f) << 16) | + ((lfi->lvl[i >> 1][1][i & 1] & 0x3f) << 8) | + (lfi->lvl[i >> 1][0][i & 1] & 0x3f); + if (!default_filt_lvl) + level = 0; + + amvdec_write_dos(core, HEVC_DBLK_CFGA, level); + } +} + +static void codec_vp9_flush_output(struct amvdec_session *sess) +{ + struct codec_vp9 *vp9 = sess->priv; + struct vp9_frame *tmp, *n; + + list_for_each_entry_safe(tmp, n, &vp9->ref_frames_list, list) { + if (!tmp->done) { + if (tmp->show) + amvdec_dst_buf_done(sess, tmp->vbuf, + V4L2_FIELD_NONE); + else + v4l2_m2m_buf_queue(sess->m2m_ctx, tmp->vbuf); + + vp9->frames_num--; + } + + list_del(&tmp->list); + kfree(tmp); + } +} + +static u32 codec_vp9_num_pending_bufs(struct amvdec_session *sess) +{ + struct codec_vp9 *vp9 = sess->priv; + + if (!vp9) + return 0; + + return vp9->frames_num; +} + +static int codec_vp9_alloc_workspace(struct amvdec_core *core, + struct codec_vp9 *vp9) +{ + /* Allocate some memory for the VP9 decoder's state */ + vp9->workspace_vaddr = dma_alloc_coherent(core->dev, SIZE_WORKSPACE, + &vp9->workspace_paddr, + GFP_KERNEL); + if (!vp9->workspace_vaddr) { + dev_err(core->dev, "Failed to allocate VP9 Workspace\n"); + return -ENOMEM; + } + + return 0; +} + +static void codec_vp9_setup_workspace(struct amvdec_session *sess, + struct codec_vp9 *vp9) +{ + struct amvdec_core *core = sess->core; + u32 revision = core->platform->revision; + dma_addr_t wkaddr = vp9->workspace_paddr; + + amvdec_write_dos(core, HEVCD_IPP_LINEBUFF_BASE, wkaddr + IPP_OFFSET); + amvdec_write_dos(core, VP9_RPM_BUFFER, wkaddr + RPM_OFFSET); + amvdec_write_dos(core, VP9_SHORT_TERM_RPS, wkaddr + SH_TM_RPS_OFFSET); + amvdec_write_dos(core, VP9_PPS_BUFFER, wkaddr + PPS_OFFSET); + amvdec_write_dos(core, VP9_SAO_UP, wkaddr + SAO_UP_OFFSET); + + amvdec_write_dos(core, VP9_STREAM_SWAP_BUFFER, + wkaddr + SWAP_BUF_OFFSET); + amvdec_write_dos(core, VP9_STREAM_SWAP_BUFFER2, + wkaddr + SWAP_BUF2_OFFSET); + amvdec_write_dos(core, VP9_SCALELUT, wkaddr + SCALELUT_OFFSET); + + if (core->platform->revision >= VDEC_REVISION_G12A) + amvdec_write_dos(core, HEVC_DBLK_CFGE, + wkaddr + DBLK_PARA_OFFSET); + + amvdec_write_dos(core, HEVC_DBLK_CFG4, wkaddr + DBLK_PARA_OFFSET); + amvdec_write_dos(core, HEVC_DBLK_CFG5, wkaddr + DBLK_DATA_OFFSET); + amvdec_write_dos(core, VP9_SEG_MAP_BUFFER, wkaddr + SEG_MAP_OFFSET); + amvdec_write_dos(core, VP9_PROB_SWAP_BUFFER, wkaddr + PROB_OFFSET); + amvdec_write_dos(core, VP9_COUNT_SWAP_BUFFER, wkaddr + COUNT_OFFSET); + amvdec_write_dos(core, LMEM_DUMP_ADR, wkaddr + LMEM_OFFSET); + + if (codec_hevc_use_mmu(revision, sess->pixfmt_cap, vp9->is_10bit)) { + amvdec_write_dos(core, HEVC_SAO_MMU_VH0_ADDR, + wkaddr + MMU_VBH_OFFSET); + amvdec_write_dos(core, HEVC_SAO_MMU_VH1_ADDR, + wkaddr + MMU_VBH_OFFSET + (MMU_VBH_SIZE / 2)); + + if (revision >= VDEC_REVISION_G12A) + amvdec_write_dos(core, HEVC_ASSIST_MMU_MAP_ADDR, + vp9->common.mmu_map_paddr); + else + amvdec_write_dos(core, VP9_MMU_MAP_BUFFER, + vp9->common.mmu_map_paddr); + } +} + +static int codec_vp9_start(struct amvdec_session *sess) +{ + struct amvdec_core *core = sess->core; + struct codec_vp9 *vp9; + u32 val; + int i; + int ret; + + vp9 = kzalloc(sizeof(*vp9), GFP_KERNEL); + if (!vp9) + return -ENOMEM; + + ret = codec_vp9_alloc_workspace(core, vp9); + if (ret) + goto free_vp9; + + codec_vp9_setup_workspace(sess, vp9); + amvdec_write_dos_bits(core, HEVC_STREAM_CONTROL, BIT(0)); + /* stream_fifo_hole */ + if (core->platform->revision >= VDEC_REVISION_G12A) + amvdec_write_dos_bits(core, HEVC_STREAM_FIFO_CTL, BIT(29)); + + val = amvdec_read_dos(core, HEVC_PARSER_INT_CONTROL) & 0x7fffffff; + val |= (3 << 29) | BIT(24) | BIT(22) | BIT(7) | BIT(4) | BIT(0); + amvdec_write_dos(core, HEVC_PARSER_INT_CONTROL, val); + amvdec_write_dos_bits(core, HEVC_SHIFT_STATUS, BIT(0)); + amvdec_write_dos(core, HEVC_SHIFT_CONTROL, BIT(10) | BIT(9) | + (3 << 6) | BIT(5) | BIT(2) | BIT(1) | BIT(0)); + amvdec_write_dos(core, HEVC_CABAC_CONTROL, BIT(0)); + amvdec_write_dos(core, HEVC_PARSER_CORE_CONTROL, BIT(0)); + amvdec_write_dos(core, HEVC_SHIFT_STARTCODE, 0x00000001); + + amvdec_write_dos(core, VP9_DEC_STATUS_REG, 0); + + amvdec_write_dos(core, HEVC_PARSER_CMD_WRITE, BIT(16)); + for (i = 0; i < ARRAY_SIZE(vdec_hevc_parser_cmd); ++i) + amvdec_write_dos(core, HEVC_PARSER_CMD_WRITE, + vdec_hevc_parser_cmd[i]); + + amvdec_write_dos(core, HEVC_PARSER_CMD_SKIP_0, PARSER_CMD_SKIP_CFG_0); + amvdec_write_dos(core, HEVC_PARSER_CMD_SKIP_1, PARSER_CMD_SKIP_CFG_1); + amvdec_write_dos(core, HEVC_PARSER_CMD_SKIP_2, PARSER_CMD_SKIP_CFG_2); + amvdec_write_dos(core, HEVC_PARSER_IF_CONTROL, + BIT(5) | BIT(2) | BIT(0)); + + amvdec_write_dos(core, HEVCD_IPP_TOP_CNTL, BIT(0)); + amvdec_write_dos(core, HEVCD_IPP_TOP_CNTL, BIT(1)); + + amvdec_write_dos(core, VP9_WAIT_FLAG, 1); + + /* clear mailbox interrupt */ + amvdec_write_dos(core, HEVC_ASSIST_MBOX1_CLR_REG, 1); + /* enable mailbox interrupt */ + amvdec_write_dos(core, HEVC_ASSIST_MBOX1_MASK, 1); + /* disable PSCALE for hardware sharing */ + amvdec_write_dos(core, HEVC_PSCALE_CTRL, 0); + /* Let the uCode do all the parsing */ + amvdec_write_dos(core, NAL_SEARCH_CTL, 0x8); + + amvdec_write_dos(core, DECODE_STOP_POS, 0); + amvdec_write_dos(core, VP9_DECODE_MODE, DECODE_MODE_SINGLE); + + pr_debug("decode_count: %u; decode_size: %u\n", + amvdec_read_dos(core, HEVC_DECODE_COUNT), + amvdec_read_dos(core, HEVC_DECODE_SIZE)); + + vp9_loop_filter_init(core, vp9); + + INIT_LIST_HEAD(&vp9->ref_frames_list); + mutex_init(&vp9->lock); + memset(&vp9->ref_frame_map, -1, sizeof(vp9->ref_frame_map)); + memset(&vp9->next_ref_frame_map, -1, sizeof(vp9->next_ref_frame_map)); + sess->priv = vp9; + + return 0; + +free_vp9: + kfree(vp9); + return ret; +} + +static int codec_vp9_stop(struct amvdec_session *sess) +{ + struct amvdec_core *core = sess->core; + struct codec_vp9 *vp9 = sess->priv; + + if (vp9->workspace_vaddr) + dma_free_coherent(core->dev, SIZE_WORKSPACE, + vp9->workspace_vaddr, + vp9->workspace_paddr); + + codec_hevc_free_fbc_buffers(sess, &vp9->common); + return 0; +} + +static void codec_vp9_set_sao(struct amvdec_session *sess, + struct vb2_buffer *vb) +{ + struct amvdec_core *core = sess->core; + struct codec_vp9 *vp9 = sess->priv; + + dma_addr_t buf_y_paddr; + dma_addr_t buf_u_v_paddr; + u32 val; + + if (codec_hevc_use_downsample(sess->pixfmt_cap, vp9->is_10bit)) + buf_y_paddr = + vp9->common.fbc_buffer_paddr[vb->index]; + else + buf_y_paddr = + vb2_dma_contig_plane_dma_addr(vb, 0); + + if (codec_hevc_use_fbc(sess->pixfmt_cap, vp9->is_10bit)) { + val = amvdec_read_dos(core, HEVC_SAO_CTRL5) & ~0xff0200; + amvdec_write_dos(core, HEVC_SAO_CTRL5, val); + amvdec_write_dos(core, HEVC_CM_BODY_START_ADDR, buf_y_paddr); + } + + if (sess->pixfmt_cap == V4L2_PIX_FMT_NV12M) { + buf_y_paddr = + vb2_dma_contig_plane_dma_addr(vb, 0); + buf_u_v_paddr = + vb2_dma_contig_plane_dma_addr(vb, 1); + amvdec_write_dos(core, HEVC_SAO_Y_START_ADDR, buf_y_paddr); + amvdec_write_dos(core, HEVC_SAO_C_START_ADDR, buf_u_v_paddr); + amvdec_write_dos(core, HEVC_SAO_Y_WPTR, buf_y_paddr); + amvdec_write_dos(core, HEVC_SAO_C_WPTR, buf_u_v_paddr); + } + + if (codec_hevc_use_mmu(core->platform->revision, sess->pixfmt_cap, + vp9->is_10bit)) { + amvdec_write_dos(core, HEVC_CM_HEADER_START_ADDR, + vp9->common.mmu_header_paddr[vb->index]); + /* use HEVC_CM_HEADER_START_ADDR */ + amvdec_write_dos_bits(core, HEVC_SAO_CTRL5, BIT(10)); + } + + amvdec_write_dos(core, HEVC_SAO_Y_LENGTH, + amvdec_get_output_size(sess)); + amvdec_write_dos(core, HEVC_SAO_C_LENGTH, + (amvdec_get_output_size(sess) / 2)); + + if (core->platform->revision >= VDEC_REVISION_G12A) { + amvdec_clear_dos_bits(core, HEVC_DBLK_CFGB, + BIT(4) | BIT(5) | BIT(8) | BIT(9)); + /* enable first, compressed write */ + if (codec_hevc_use_fbc(sess->pixfmt_cap, vp9->is_10bit)) + amvdec_write_dos_bits(core, HEVC_DBLK_CFGB, BIT(8)); + + /* enable second, uncompressed write */ + if (sess->pixfmt_cap == V4L2_PIX_FMT_NV12M) + amvdec_write_dos_bits(core, HEVC_DBLK_CFGB, BIT(9)); + + /* dblk pipeline mode=1 for performance */ + if (sess->width >= 1280) + amvdec_write_dos_bits(core, HEVC_DBLK_CFGB, BIT(4)); + + pr_debug("HEVC_DBLK_CFGB: %08X\n", + amvdec_read_dos(core, HEVC_DBLK_CFGB)); + } + + val = amvdec_read_dos(core, HEVC_SAO_CTRL1) & ~0x3ff0; + val |= 0xff0; /* Set endianness for 2-bytes swaps (nv12) */ + if (core->platform->revision < VDEC_REVISION_G12A) { + val &= ~0x3; + if (!codec_hevc_use_fbc(sess->pixfmt_cap, vp9->is_10bit)) + val |= BIT(0); /* disable cm compression */ + /* TOFIX: Handle Amlogic Framebuffer compression */ + } + + amvdec_write_dos(core, HEVC_SAO_CTRL1, val); + pr_debug("HEVC_SAO_CTRL1: %08X\n", val); + + /* no downscale for NV12 */ + val = amvdec_read_dos(core, HEVC_SAO_CTRL5) & ~0xff0000; + amvdec_write_dos(core, HEVC_SAO_CTRL5, val); + + val = amvdec_read_dos(core, HEVCD_IPP_AXIIF_CONFIG) & ~0x30; + val |= 0xf; + val &= ~BIT(12); /* NV12 */ + amvdec_write_dos(core, HEVCD_IPP_AXIIF_CONFIG, val); +} + +static dma_addr_t codec_vp9_get_frame_mv_paddr(struct codec_vp9 *vp9, + struct vp9_frame *frame) +{ + return vp9->workspace_paddr + MPRED_MV_OFFSET + + (frame->index * MPRED_MV_BUF_SIZE); +} + +static void codec_vp9_set_mpred_mv(struct amvdec_core *core, + struct codec_vp9 *vp9) +{ + int mpred_mv_rd_end_addr; + int use_prev_frame_mvs = !vp9->prev_frame->intra_only && + vp9->prev_frame->show && + vp9->prev_frame->type != KEY_FRAME; + + amvdec_write_dos(core, HEVC_MPRED_CTRL3, 0x24122412); + amvdec_write_dos(core, HEVC_MPRED_ABV_START_ADDR, + vp9->workspace_paddr + MPRED_ABV_OFFSET); + + amvdec_clear_dos_bits(core, HEVC_MPRED_CTRL4, BIT(6)); + if (use_prev_frame_mvs) + amvdec_write_dos_bits(core, HEVC_MPRED_CTRL4, BIT(6)); + + amvdec_write_dos(core, HEVC_MPRED_MV_WR_START_ADDR, + codec_vp9_get_frame_mv_paddr(vp9, vp9->cur_frame)); + amvdec_write_dos(core, HEVC_MPRED_MV_WPTR, + codec_vp9_get_frame_mv_paddr(vp9, vp9->cur_frame)); + + amvdec_write_dos(core, HEVC_MPRED_MV_RD_START_ADDR, + codec_vp9_get_frame_mv_paddr(vp9, vp9->prev_frame)); + amvdec_write_dos(core, HEVC_MPRED_MV_RPTR, + codec_vp9_get_frame_mv_paddr(vp9, vp9->prev_frame)); + + mpred_mv_rd_end_addr = + codec_vp9_get_frame_mv_paddr(vp9, vp9->prev_frame) + + (vp9->lcu_total * MV_MEM_UNIT); + amvdec_write_dos(core, HEVC_MPRED_MV_RD_END_ADDR, mpred_mv_rd_end_addr); +} + +static void codec_vp9_update_next_ref(struct codec_vp9 *vp9) +{ + union rpm_param *param = &vp9->rpm_param; + u32 buf_idx = vp9->cur_frame->index; + int ref_index = 0; + int refresh_frame_flags; + int mask; + + refresh_frame_flags = vp9->cur_frame->type == KEY_FRAME ? + 0xff : param->p.refresh_frame_flags; + + for (mask = refresh_frame_flags; mask; mask >>= 1) { + pr_debug("mask=%08X; ref_index=%d\n", mask, ref_index); + if (mask & 1) + vp9->next_ref_frame_map[ref_index] = buf_idx; + else + vp9->next_ref_frame_map[ref_index] = + vp9->ref_frame_map[ref_index]; + + ++ref_index; + } + + for (; ref_index < REF_FRAMES; ++ref_index) + vp9->next_ref_frame_map[ref_index] = + vp9->ref_frame_map[ref_index]; +} + +static void codec_vp9_update_ref(struct codec_vp9 *vp9) +{ + union rpm_param *param = &vp9->rpm_param; + int ref_index = 0; + int mask; + int refresh_frame_flags; + + if (!vp9->cur_frame) + return; + + refresh_frame_flags = vp9->cur_frame->type == KEY_FRAME ? + 0xff : param->p.refresh_frame_flags; + + for (mask = refresh_frame_flags; mask; mask >>= 1) { + vp9->ref_frame_map[ref_index] = + vp9->next_ref_frame_map[ref_index]; + ++ref_index; + } + + if (param->p.show_existing_frame) + return; + + for (; ref_index < REF_FRAMES; ++ref_index) + vp9->ref_frame_map[ref_index] = + vp9->next_ref_frame_map[ref_index]; +} + +static struct vp9_frame *codec_vp9_get_frame_by_idx(struct codec_vp9 *vp9, + int idx) +{ + struct vp9_frame *frame; + + list_for_each_entry(frame, &vp9->ref_frames_list, list) { + if (frame->index == idx) + return frame; + } + + return NULL; +} + +static void codec_vp9_sync_ref(struct codec_vp9 *vp9) +{ + union rpm_param *param = &vp9->rpm_param; + int i; + + for (i = 0; i < REFS_PER_FRAME; ++i) { + const int ref = + (param->p.ref_info >> + (((REFS_PER_FRAME - i - 1) * 4) + 1)) & 0x7; + const int idx = vp9->ref_frame_map[ref]; + + vp9->frame_refs[i] = codec_vp9_get_frame_by_idx(vp9, idx); + } +} + +static void codec_vp9_set_refs(struct amvdec_session *sess, + struct codec_vp9 *vp9) +{ + struct amvdec_core *core = sess->core; + int i; + + for (i = 0; i < REFS_PER_FRAME; ++i) { + struct vp9_frame *frame = vp9->frame_refs[i]; + int id_y; + int id_u_v; + + if (!frame) + continue; + + if (codec_hevc_use_fbc(sess->pixfmt_cap, vp9->is_10bit)) { + id_y = frame->index; + id_u_v = id_y; + } else { + id_y = frame->index * 2; + id_u_v = id_y + 1; + } + + amvdec_write_dos(core, HEVCD_MPP_ANC_CANVAS_DATA_ADDR, + (id_u_v << 16) | (id_u_v << 8) | id_y); + } +} + +static void codec_vp9_set_mc(struct amvdec_session *sess, + struct codec_vp9 *vp9) +{ + struct amvdec_core *core = sess->core; + int i; + + amvdec_write_dos(core, HEVCD_MPP_ANC_CANVAS_ACCCONFIG_ADDR, 1); + codec_vp9_set_refs(sess, vp9); + amvdec_write_dos(core, HEVCD_MPP_ANC_CANVAS_ACCCONFIG_ADDR, + (16 << 8) | 1); + codec_vp9_set_refs(sess, vp9); + + amvdec_write_dos(core, VP9D_MPP_REFINFO_TBL_ACCCONFIG, BIT(2)); + for (i = 0; i < REFS_PER_FRAME; ++i) { + if (!vp9->frame_refs[i]) + continue; + + amvdec_write_dos(core, VP9D_MPP_REFINFO_DATA, vp9->width); + amvdec_write_dos(core, VP9D_MPP_REFINFO_DATA, vp9->height); + amvdec_write_dos(core, VP9D_MPP_REFINFO_DATA, + (vp9->width << 14) / vp9->width); + amvdec_write_dos(core, VP9D_MPP_REFINFO_DATA, + (vp9->height << 14) / vp9->height); + amvdec_write_dos(core, VP9D_MPP_REFINFO_DATA, + amvdec_am21c_body_size(vp9->width, + vp9->height) >> 5); + } + + amvdec_write_dos(core, VP9D_MPP_REF_SCALE_ENBL, 0); +} + +static struct vp9_frame *codec_vp9_get_new_frame(struct amvdec_session *sess) +{ + struct codec_vp9 *vp9 = sess->priv; + union rpm_param *param = &vp9->rpm_param; + struct vb2_v4l2_buffer *vbuf; + struct vp9_frame *new_frame; + + new_frame = kzalloc(sizeof(*new_frame), GFP_KERNEL); + if (!new_frame) + return NULL; + + vbuf = v4l2_m2m_dst_buf_remove(sess->m2m_ctx); + if (!vbuf) { + dev_err(sess->core->dev, "No dst buffer available\n"); + kfree(new_frame); + return NULL; + } + + while (codec_vp9_get_frame_by_idx(vp9, vbuf->vb2_buf.index)) { + struct vb2_v4l2_buffer *old_vbuf = vbuf; + + vbuf = v4l2_m2m_dst_buf_remove(sess->m2m_ctx); + v4l2_m2m_buf_queue(sess->m2m_ctx, old_vbuf); + if (!vbuf) { + dev_err(sess->core->dev, "No dst buffer available\n"); + kfree(new_frame); + return NULL; + } + } + + new_frame->vbuf = vbuf; + new_frame->index = vbuf->vb2_buf.index; + new_frame->intra_only = param->p.intra_only; + new_frame->show = param->p.show_frame; + new_frame->type = param->p.frame_type; + list_add_tail(&new_frame->list, &vp9->ref_frames_list); + vp9->frames_num++; + + return new_frame; +} + +static void codec_vp9_show_existing_frame(struct codec_vp9 *vp9) +{ + union rpm_param *param = &vp9->rpm_param; + + if (!param->p.show_existing_frame) + return; + + pr_debug("showing frame %u\n", param->p.frame_to_show_idx); +} + +static void codec_vp9_rm_noshow_frame(struct amvdec_session *sess) +{ + struct codec_vp9 *vp9 = sess->priv; + struct vp9_frame *tmp; + + list_for_each_entry(tmp, &vp9->ref_frames_list, list) { + if (tmp->show) + continue; + + pr_debug("rm noshow: %u\n", tmp->index); + v4l2_m2m_buf_queue(sess->m2m_ctx, tmp->vbuf); + list_del(&tmp->list); + kfree(tmp); + vp9->frames_num--; + return; + } +} + +static void codec_vp9_process_frame(struct amvdec_session *sess) +{ + struct amvdec_core *core = sess->core; + struct codec_vp9 *vp9 = sess->priv; + union rpm_param *param = &vp9->rpm_param; + int intra_only; + + if (!param->p.show_frame) + codec_vp9_rm_noshow_frame(sess); + + vp9->cur_frame = codec_vp9_get_new_frame(sess); + if (!vp9->cur_frame) + return; + + pr_debug("frame type: %08X; show_exist: %u; show: %u, intra_only: %u\n", + param->p.frame_type, param->p.show_existing_frame, + param->p.show_frame, param->p.intra_only); + + codec_vp9_sync_ref(vp9); + codec_vp9_update_next_ref(vp9); + codec_vp9_show_existing_frame(vp9); + + if (codec_hevc_use_mmu(core->platform->revision, sess->pixfmt_cap, + vp9->is_10bit)) + codec_hevc_fill_mmu_map(sess, &vp9->common, + &vp9->cur_frame->vbuf->vb2_buf); + + intra_only = param->p.show_frame ? 0 : param->p.intra_only; + + /* clear mpred (for keyframe only) */ + if (param->p.frame_type != KEY_FRAME && !intra_only) { + codec_vp9_set_mc(sess, vp9); + codec_vp9_set_mpred_mv(core, vp9); + } else { + amvdec_clear_dos_bits(core, HEVC_MPRED_CTRL4, BIT(6)); + } + + amvdec_write_dos(core, HEVC_PARSER_PICTURE_SIZE, + (vp9->height << 16) | vp9->width); + codec_vp9_set_sao(sess, &vp9->cur_frame->vbuf->vb2_buf); + + vp9_loop_filter_frame_init(core, &vp9->seg_4lf, + &vp9->lfi, &vp9->lf, + vp9->default_filt_lvl); + + /* ask uCode to start decoding */ + amvdec_write_dos(core, VP9_DEC_STATUS_REG, VP9_10B_DECODE_SLICE); +} + +static void codec_vp9_process_lf(struct codec_vp9 *vp9) +{ + union rpm_param *param = &vp9->rpm_param; + int i; + + vp9->lf.mode_ref_delta_enabled = param->p.mode_ref_delta_enabled; + vp9->lf.sharpness_level = param->p.sharpness_level; + vp9->default_filt_lvl = param->p.filter_level; + vp9->seg_4lf.enabled = param->p.seg_enabled; + vp9->seg_4lf.abs_delta = param->p.seg_abs_delta; + + for (i = 0; i < 4; i++) + vp9->lf.ref_deltas[i] = param->p.ref_deltas[i]; + + for (i = 0; i < 2; i++) + vp9->lf.mode_deltas[i] = param->p.mode_deltas[i]; + + for (i = 0; i < MAX_SEGMENTS; i++) + vp9->seg_4lf.feature_mask[i] = + (param->p.seg_lf_info[i] & 0x8000) ? + (1 << SEG_LVL_ALT_LF) : 0; + + for (i = 0; i < MAX_SEGMENTS; i++) + vp9->seg_4lf.feature_data[i][SEG_LVL_ALT_LF] = + (param->p.seg_lf_info[i] & 0x100) ? + -(param->p.seg_lf_info[i] & 0x3f) + : (param->p.seg_lf_info[i] & 0x3f); +} + +static void codec_vp9_resume(struct amvdec_session *sess) +{ + struct codec_vp9 *vp9 = sess->priv; + + if (codec_hevc_setup_buffers(sess, &vp9->common, vp9->is_10bit)) { + amvdec_abort(sess); + return; + } + + codec_vp9_setup_workspace(sess, vp9); + codec_hevc_setup_decode_head(sess, vp9->is_10bit); + codec_vp9_process_lf(vp9); + codec_vp9_process_frame(sess); +} + +/** + * The RPM section within the workspace contains + * many information regarding the parsed bitstream + */ +static void codec_vp9_fetch_rpm(struct amvdec_session *sess) +{ + struct codec_vp9 *vp9 = sess->priv; + u16 *rpm_vaddr = vp9->workspace_vaddr + RPM_OFFSET; + int i, j; + + for (i = 0; i < RPM_BUF_SIZE; i += 4) + for (j = 0; j < 4; j++) + vp9->rpm_param.l.data[i + j] = rpm_vaddr[i + 3 - j]; +} + +static int codec_vp9_process_rpm(struct codec_vp9 *vp9) +{ + union rpm_param *param = &vp9->rpm_param; + int src_changed = 0; + int is_10bit = 0; + int pic_width_64 = ALIGN(param->p.width, 64); + int pic_height_32 = ALIGN(param->p.height, 32); + int pic_width_lcu = (pic_width_64 % LCU_SIZE) ? + pic_width_64 / LCU_SIZE + 1 + : pic_width_64 / LCU_SIZE; + int pic_height_lcu = (pic_height_32 % LCU_SIZE) ? + pic_height_32 / LCU_SIZE + 1 + : pic_height_32 / LCU_SIZE; + vp9->lcu_total = pic_width_lcu * pic_height_lcu; + + if (param->p.bit_depth == 10) + is_10bit = 1; + + if (vp9->width != param->p.width || vp9->height != param->p.height || + vp9->is_10bit != is_10bit) + src_changed = 1; + + vp9->width = param->p.width; + vp9->height = param->p.height; + vp9->is_10bit = is_10bit; + + pr_debug("width: %u; height: %u; is_10bit: %d; src_changed: %d\n", + vp9->width, vp9->height, is_10bit, src_changed); + + return src_changed; +} + +static bool codec_vp9_is_ref(struct codec_vp9 *vp9, struct vp9_frame *frame) +{ + int i; + + for (i = 0; i < REFS_PER_FRAME; ++i) + if (vp9->frame_refs[i] == frame) + return true; + + return false; +} + +static void codec_vp9_show_frame(struct amvdec_session *sess) +{ + struct codec_vp9 *vp9 = sess->priv; + struct vp9_frame *tmp, *n; + + list_for_each_entry_safe(tmp, n, &vp9->ref_frames_list, list) { + if (!tmp->show || tmp == vp9->cur_frame) + continue; + + if (!tmp->done) { + pr_debug("Doning %u\n", tmp->index); + amvdec_dst_buf_done(sess, tmp->vbuf, V4L2_FIELD_NONE); + tmp->done = 1; + vp9->frames_num--; + } + + if (codec_vp9_is_ref(vp9, tmp)) + continue; + + pr_debug("deleting %d\n", tmp->index); + list_del(&tmp->list); + kfree(tmp); + } +} + +static irqreturn_t codec_vp9_threaded_isr(struct amvdec_session *sess) +{ + struct amvdec_core *core = sess->core; + struct codec_vp9 *vp9 = sess->priv; + u32 dec_status = amvdec_read_dos(core, VP9_DEC_STATUS_REG); + u32 prob_status = amvdec_read_dos(core, VP9_ADAPT_PROB_REG); + int i; + + if (!vp9) + return IRQ_HANDLED; + + mutex_lock(&vp9->lock); + if (dec_status != VP9_HEAD_PARSER_DONE) { + dev_err(core->dev_dec, "Unrecognized dec_status: %08X\n", + dec_status); + amvdec_abort(sess); + goto unlock; + } + + pr_debug("ISR: %08X;%08X\n", dec_status, prob_status); + sess->keyframe_found = 1; + + /* Invalidate first 3 refs */ + for (i = 0; i < 3; ++i) + vp9->frame_refs[i] = NULL; + + vp9->prev_frame = vp9->cur_frame; + codec_vp9_update_ref(vp9); + + codec_vp9_fetch_rpm(sess); + if (codec_vp9_process_rpm(vp9)) { + amvdec_src_change(sess, vp9->width, vp9->height, 16); + goto unlock; + } + + codec_vp9_process_lf(vp9); + codec_vp9_process_frame(sess); + codec_vp9_show_frame(sess); + +unlock: + mutex_unlock(&vp9->lock); + return IRQ_HANDLED; +} + +static irqreturn_t codec_vp9_isr(struct amvdec_session *sess) +{ + return IRQ_WAKE_THREAD; +} + +struct amvdec_codec_ops codec_vp9_ops = { + .start = codec_vp9_start, + .stop = codec_vp9_stop, + .isr = codec_vp9_isr, + .threaded_isr = codec_vp9_threaded_isr, + .num_pending_bufs = codec_vp9_num_pending_bufs, + .drain = codec_vp9_flush_output, + .resume = codec_vp9_resume, +}; diff --git a/drivers/staging/media/meson/vdec/codec_vp9.h b/drivers/staging/media/meson/vdec/codec_vp9.h new file mode 100644 index 000000000000..62db65a2b939 --- /dev/null +++ b/drivers/staging/media/meson/vdec/codec_vp9.h @@ -0,0 +1,13 @@ +/* SPDX-License-Identifier: GPL-2.0+ */ +/* + * Copyright (C) 2018 Maxime Jourdan + */ + +#ifndef __MESON_VDEC_CODEC_VP9_H_ +#define __MESON_VDEC_CODEC_VP9_H_ + +#include "vdec.h" + +extern struct amvdec_codec_ops codec_vp9_ops; + +#endif diff --git a/drivers/staging/media/meson/vdec/hevc_regs.h b/drivers/staging/media/meson/vdec/hevc_regs.h index 55c1a80b955a..0392f41a1eed 100644 --- a/drivers/staging/media/meson/vdec/hevc_regs.h +++ b/drivers/staging/media/meson/vdec/hevc_regs.h @@ -122,6 +122,8 @@ #define HEVC_MPRED_L0_REF00_POC 0xc880 #define HEVC_MPRED_L1_REF00_POC 0xc8c0 +#define HEVC_MPRED_CTRL4 0xc930 + #define HEVC_MPRED_CUR_POC 0xc980 #define HEVC_MPRED_COL_POC 0xc984 #define HEVC_MPRED_MV_RD_END_ADDR 0xc988 @@ -140,6 +142,10 @@ #define HEVCD_IPP_LINEBUFF_BASE 0xd024 #define HEVCD_IPP_AXIIF_CONFIG 0xd02c +#define VP9D_MPP_REF_SCALE_ENBL 0xd104 +#define VP9D_MPP_REFINFO_TBL_ACCCONFIG 0xd108 +#define VP9D_MPP_REFINFO_DATA 0xd10c + #define HEVCD_MPP_ANC2AXI_TBL_CONF_ADDR 0xd180 #define HEVCD_MPP_ANC2AXI_TBL_CMD_ADDR 0xd184 #define HEVCD_MPP_ANC2AXI_TBL_DATA 0xd190 @@ -164,6 +170,7 @@ #define HEVC_DBLK_CFG9 0xd424 #define HEVC_DBLK_CFGA 0xd428 #define HEVC_DBLK_STS0 0xd42c +#define HEVC_DBLK_CFGB 0xd42c #define HEVC_DBLK_STS1 0xd430 #define HEVC_DBLK_CFGE 0xd438 diff --git a/drivers/staging/media/meson/vdec/vdec_platform.c b/drivers/staging/media/meson/vdec/vdec_platform.c index e9356a46828f..72a833b1cebd 100644 --- a/drivers/staging/media/meson/vdec/vdec_platform.c +++ b/drivers/staging/media/meson/vdec/vdec_platform.c @@ -8,8 +8,10 @@ #include "vdec.h" #include "vdec_1.h" +#include "vdec_hevc.h" #include "codec_mpeg12.h" #include "codec_h264.h" +#include "codec_vp9.h" static const struct amvdec_format vdec_formats_gxbb[] = { { @@ -51,6 +53,18 @@ static const struct amvdec_format vdec_formats_gxbb[] = { static const struct amvdec_format vdec_formats_gxl[] = { { + .pixfmt = V4L2_PIX_FMT_VP9, + .min_buffers = 16, + .max_buffers = 24, + .max_width = 3840, + .max_height = 2160, + .vdec_ops = &vdec_hevc_ops, + .codec_ops = &codec_vp9_ops, + .firmware_path = "meson/vdec/gxl_vp9.bin", + .pixfmts_cap = { V4L2_PIX_FMT_NV12M, 0 }, + .flags = V4L2_FMT_FLAG_COMPRESSED | + V4L2_FMT_FLAG_DYN_RESOLUTION, + }, { .pixfmt = V4L2_PIX_FMT_H264, .min_buffers = 2, .max_buffers = 24, @@ -127,6 +141,18 @@ static const struct amvdec_format vdec_formats_gxm[] = { static const struct amvdec_format vdec_formats_g12a[] = { { + .pixfmt = V4L2_PIX_FMT_VP9, + .min_buffers = 16, + .max_buffers = 24, + .max_width = 3840, + .max_height = 2160, + .vdec_ops = &vdec_hevc_ops, + .codec_ops = &codec_vp9_ops, + .firmware_path = "meson/vdec/g12a_vp9.bin", + .pixfmts_cap = { V4L2_PIX_FMT_NV12M, 0 }, + .flags = V4L2_FMT_FLAG_COMPRESSED | + V4L2_FMT_FLAG_DYN_RESOLUTION, + }, { .pixfmt = V4L2_PIX_FMT_H264, .min_buffers = 2, .max_buffers = 24, @@ -165,6 +191,18 @@ static const struct amvdec_format vdec_formats_g12a[] = { static const struct amvdec_format vdec_formats_sm1[] = { { + .pixfmt = V4L2_PIX_FMT_VP9, + .min_buffers = 16, + .max_buffers = 24, + .max_width = 3840, + .max_height = 2160, + .vdec_ops = &vdec_hevc_ops, + .codec_ops = &codec_vp9_ops, + .firmware_path = "meson/vdec/g12a_vp9.bin", + .pixfmts_cap = { V4L2_PIX_FMT_NV12M, 0 }, + .flags = V4L2_FMT_FLAG_COMPRESSED | + V4L2_FMT_FLAG_DYN_RESOLUTION, + }, { .pixfmt = V4L2_PIX_FMT_H264, .min_buffers = 2, .max_buffers = 24,