From mboxrd@z Thu Jan 1 00:00:00 1970 Return-path: Received: from mail-bk0-f46.google.com ([209.85.214.46]:33078 "EHLO mail-bk0-f46.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754362Ab2IWUld (ORCPT ); Sun, 23 Sep 2012 16:41:33 -0400 Received: by bkcjk13 with SMTP id jk13so562504bkc.19 for ; Sun, 23 Sep 2012 13:41:31 -0700 (PDT) Message-ID: <505F65E8.6040601@googlemail.com> Date: Sun, 23 Sep 2012 22:41:28 +0300 From: =?UTF-8?B?RnJhbmsgU2Now6RmZXI=?= MIME-Version: 1.0 To: Guennadi Liakhovetski CC: maramaopercheseimorto@gmail.com, linux-media@vger.kernel.org Subject: Re: [PATCH v2 1/3] ov2640: select sensor register bank before applying h/v-flip settings References: <1348424926-12864-1-git-send-email-fschaefer.oss@googlemail.com> In-Reply-To: Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Sender: linux-media-owner@vger.kernel.org List-ID: Am 23.09.2012 23:21, schrieb Guennadi Liakhovetski: > Hi Frank > > On Sun, 23 Sep 2012, Frank Schäfer wrote: > >> We currently don't select the register bank in ov2640_s_ctrl, so we can end up >> writing to DSP register 0x04 instead of sensor register 0x04. >> This happens for example when calling ov2640_s_ctrl after ov2640_s_fmt. > Yes, in principle, I agree, bank switching in the driver is not very... > consistent and also this specific case looks buggy. But, we have to fix > your fix. > >> Signed-off-by: Frank Schäfer >> Cc: stable@kernel.org >> --- >> drivers/media/i2c/soc_camera/ov2640.c | 8 ++++++++ >> 1 Datei geändert, 8 Zeilen hinzugefügt(+) >> >> diff --git a/drivers/media/i2c/soc_camera/ov2640.c b/drivers/media/i2c/soc_camera/ov2640.c >> index 78ac574..e4fc79e 100644 >> --- a/drivers/media/i2c/soc_camera/ov2640.c >> +++ b/drivers/media/i2c/soc_camera/ov2640.c >> @@ -683,8 +683,16 @@ static int ov2640_s_ctrl(struct v4l2_ctrl *ctrl) >> struct v4l2_subdev *sd = >> &container_of(ctrl->handler, struct ov2640_priv, hdl)->subdev; >> struct i2c_client *client = v4l2_get_subdevdata(sd); >> + struct regval_list regval; >> + int ret; >> u8 val; >> >> + regval.reg_num = BANK_SEL; >> + regval.value = BANK_SEL_SENS; >> + ret = ov2640_write_array(client, ®val); > This doesn't look right to me. ov2640_write_array() keeps writing register > address - value pairs to the hardware until it encounters an "ENDMARKER," > which you don't have here, so, it's hard to say what will be written to > the sensor... Secondly, you only have to write a single register here, for > this the driver is already using i2c_smbus_write_byte_data() directly, > please, do the same. Argh, yes, you're right. The mistake was to split this off from patch 3 to reduce changes for stable... I will combine both patches and resend the series. Regards, Frank > > Thanks > Guennadi > >> + if (ret < 0) >> + return ret; >> + >> switch (ctrl->id) { >> case V4L2_CID_VFLIP: >> val = ctrl->val ? REG04_VFLIP_IMG : 0x00; >> -- >> 1.7.10.4 >> > --- > Guennadi Liakhovetski, Ph.D. > Freelance Open-Source Software Developer > http://www.open-technology.de/