linux-media.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Hans Verkuil <hverkuil@xs4all.nl>
To: Dave Stevenson <dave.stevenson@raspberrypi.com>
Cc: "Laurent Pinchart" <laurent.pinchart@ideasonboard.com>,
	"Linux Media Mailing List" <linux-media@vger.kernel.org>,
	"Kieran Bingham" <kieran.bingham@ideasonboard.com>,
	"Jacopo Mondi" <jacopo@jmondi.org>,
	"Niklas Söderlund" <niklas.soderlund@ragnatech.se>,
	"Naushir Patuck" <naush@raspberrypi.com>
Subject: Re: [PATCH v2 33/34] staging: bcm2835-isp: Add support for BC2835 ISP
Date: Mon, 18 May 2020 17:07:20 +0200	[thread overview]
Message-ID: <bdcb471b-e066-373f-aa16-b7caf5c52795@xs4all.nl> (raw)
In-Reply-To: <CAPY8ntBDss5ZKPo3mRUP0-9KNHA6kQEQwrnvyUpOh0Ru7O5CKA@mail.gmail.com>

On 18/05/2020 16:36, Dave Stevenson wrote:
> Hi Hans.
> 
> Thanks for the review.
> A few of these I need to leave for Naush to answer.
> Ideas On Board have been contracted to drive this upstreaming process,
> so many of these will be left to them to address and generate a v2.

<snip>

>>> +static int bcm2835_isp_node_s_selection(struct file *file, void *fh,
>>> +                                     struct v4l2_selection *s)
>>> +{
>>> +     struct mmal_parameter_crop crop;
>>> +     struct bcm2835_isp_node *node = video_drvdata(file);
>>> +     struct bcm2835_isp_dev *dev = node_get_dev(node);
>>> +     struct vchiq_mmal_port *port = get_port_data(node);
>>> +
>>> +     /* This return value is required fro V4L2 compliance. */
>>> +     if (node_is_stats(node))
>>> +             return -ENOTTY;
>>> +
>>> +     if (!s->r.width || !s->r.height)
>>> +             return -EINVAL;
>>
>> I'm missing a check for s->target.
> 
> Ack
> 
>>> +
>>> +     /* Adjust the crop window if goes outside the frame dimensions. */
>>> +     s->r.left = min((unsigned int)max(s->r.left, 0),
>>> +                     node->q_data.width - MIN_DIM);
>>> +     s->r.top = min((unsigned int)max(s->r.top, 0),
>>> +                    node->q_data.height - MIN_DIM);
>>> +     s->r.width = max(min(s->r.width, node->q_data.width - s->r.left),
>>> +                      MIN_DIM);
>>> +     s->r.height = max(min(s->r.height, node->q_data.height - s->r.top),
>>> +                       MIN_DIM);
>>> +
>>> +     crop.rect.x = s->r.left;
>>> +     crop.rect.y = s->r.top;
>>> +     crop.rect.width = s->r.width;
>>> +     crop.rect.height = s->r.height;
>>> +
>>> +     return vchiq_mmal_port_parameter_set(dev->mmal_instance, port,
>>> +                                          MMAL_PARAMETER_CROP,
>>> +                                          &crop, sizeof(crop));
>>> +}
>>> +
>>> +static int bcm2835_isp_node_g_selection(struct file *file, void *fh,
>>> +                                     struct v4l2_selection *s)
>>> +{
>>> +     struct bcm2835_isp_node *node = video_drvdata(file);
>>> +     struct bcm2835_isp_dev *dev = node_get_dev(node);
>>> +     struct vchiq_mmal_port *port = get_port_data(node);
>>> +     struct mmal_parameter_crop crop;
>>> +     u32 crop_size = sizeof(crop);
>>> +     int ret;
>>> +
>>> +     /* This return value is required for V4L2 compliance. */
>>> +     if (node_is_stats(node))
>>> +             return -ENOTTY;
>>> +
>>> +     /* We can only return out an input crop. */
>>> +     if (s->target != V4L2_SEL_TGT_CROP)
>>> +             return -EINVAL;
>>
>> No support for _TGT_CROP_DEFAULT/BOUNDS? Those are usually supported
>> by drivers and are typically set to the width/height of the current
>> format.
>>
>> I recommend adding support for these targets.
> 
> Trying to find any good M2M drivers to use as an example was tricky,
> but if the return value can be as simple as the current width/height,
> then that's easy.
> 
>>> +
>>> +     ret = vchiq_mmal_port_parameter_get(dev->mmal_instance, port,
>>> +                                         MMAL_PARAMETER_CROP,
>>> +                                         &crop, &crop_size);
>>> +     if (!ret)
>>> +             return -EINVAL;
>>> +
>>> +     s->r.left = crop.rect.x;
>>> +     s->r.top = crop.rect.y;
>>> +     s->r.width = crop.rect.width;
>>> +     s->r.height = crop.rect.height;
>>> +
>>> +     return 0;
>>> +}
>>> +
>>> +static int bcm3285_isp_subscribe_event(struct v4l2_fh *fh,
>>> +                                    const struct v4l2_event_subscription *s)
>>> +{
>>> +     switch (s->type) {
>>> +     /* Cannot change source parameters dynamically at runtime. */
>>> +     case V4L2_EVENT_SOURCE_CHANGE:
>>> +             return -EINVAL;
>>> +     case V4L2_EVENT_CTRL:
>>> +             return v4l2_ctrl_subscribe_event(fh, s);
>>> +     default:
>>> +             return v4l2_event_subscribe(fh, s, 4, NULL);
>>> +     }
>>> +}
>>> +
>>> +static const struct v4l2_ioctl_ops bcm2835_isp_node_ioctl_ops = {
>>> +     .vidioc_querycap                = bcm2835_isp_node_querycap,
>>> +     .vidioc_g_fmt_vid_cap           = bcm2835_isp_node_g_fmt,
>>> +     .vidioc_g_fmt_vid_out           = bcm2835_isp_node_g_fmt,
>>> +     .vidioc_g_fmt_meta_cap          = bcm2835_isp_node_g_fmt,
>>> +     .vidioc_s_fmt_vid_cap           = bcm2835_isp_node_s_fmt,
>>> +     .vidioc_s_fmt_vid_out           = bcm2835_isp_node_s_fmt,
>>> +     .vidioc_s_fmt_meta_cap          = bcm2835_isp_node_s_fmt,
>>> +     .vidioc_try_fmt_vid_cap         = bcm2835_isp_node_try_fmt,
>>> +     .vidioc_try_fmt_vid_out         = bcm2835_isp_node_try_fmt,
>>> +     .vidioc_try_fmt_meta_cap        = bcm2835_isp_node_try_fmt,
>>> +     .vidioc_s_selection             = bcm2835_isp_node_s_selection,
>>> +     .vidioc_g_selection             = bcm2835_isp_node_g_selection,
>>> +
>>> +     .vidioc_enum_fmt_vid_cap        = bcm2835_isp_node_enum_fmt,
>>> +     .vidioc_enum_fmt_vid_out        = bcm2835_isp_node_enum_fmt,
>>> +     .vidioc_enum_fmt_meta_cap       = bcm2835_isp_node_enum_fmt,
>>> +     .vidioc_enum_framesizes         = bcm2835_isp_enum_framesizes,
>>> +
>>> +     .vidioc_reqbufs                 = vb2_ioctl_reqbufs,
>>> +     .vidioc_querybuf                = vb2_ioctl_querybuf,
>>> +     .vidioc_qbuf                    = vb2_ioctl_qbuf,
>>> +     .vidioc_dqbuf                   = vb2_ioctl_dqbuf,
>>> +     .vidioc_expbuf                  = vb2_ioctl_expbuf,
>>> +     .vidioc_create_bufs             = vb2_ioctl_create_bufs,
>>> +     .vidioc_prepare_buf             = vb2_ioctl_prepare_buf,
>>> +
>>> +     .vidioc_streamon                = vb2_ioctl_streamon,
>>> +     .vidioc_streamoff               = vb2_ioctl_streamoff,
>>> +
>>> +     .vidioc_subscribe_event         = bcm3285_isp_subscribe_event,
>>> +     .vidioc_unsubscribe_event       = v4l2_event_unsubscribe,
>>> +};
>>> +
>>> +/*
>>> + * Size of the array to provide to the VPU when asking for the list of supported
>>> + * formats.
>>> + *
>>> + * The ISP component currently advertises 33 input formats, so add a small
>>> + * overhead on that.
>>> + */
>>> +#define MAX_SUPPORTED_ENCODINGS 40
>>> +
>>> +/* Populate node->supported_fmts with the formats supported by those ports. */
>>> +static int bcm2835_isp_get_supported_fmts(struct bcm2835_isp_node *node)
>>> +{
>>> +     struct bcm2835_isp_dev *dev = node_get_dev(node);
>>> +     struct bcm2835_isp_fmt const **list;
>>> +     unsigned int i, j, num_encodings;
>>> +     u32 fourccs[MAX_SUPPORTED_ENCODINGS];
>>> +     u32 param_size = sizeof(fourccs);
>>> +     int ret;
>>> +
>>> +     ret = vchiq_mmal_port_parameter_get(dev->mmal_instance,
>>> +                                         get_port_data(node),
>>> +                                         MMAL_PARAMETER_SUPPORTED_ENCODINGS,
>>> +                                         &fourccs, &param_size);
>>> +
>>> +     if (ret) {
>>> +             if (ret == MMAL_MSG_STATUS_ENOSPC) {
>>> +                     v4l2_err(&dev->v4l2_dev,
>>> +                              "%s: port has more encoding than we provided space for. Some are dropped.\n",
>>> +                              __func__);
>>> +                     num_encodings = MAX_SUPPORTED_ENCODINGS;
>>> +             } else {
>>> +                     v4l2_err(&dev->v4l2_dev, "%s: get_param ret %u.\n",
>>> +                              __func__, ret);
>>> +                     return -EINVAL;
>>> +             }
>>> +     } else {
>>> +             num_encodings = param_size / sizeof(u32);
>>> +     }
>>> +
>>> +     /*
>>> +      * Assume at this stage that all encodings will be supported in V4L2.
>>> +      * Any that aren't supported will waste a very small amount of memory.
>>> +      */
>>> +     list = devm_kzalloc(dev->dev,
>>> +                         sizeof(struct bcm2835_isp_fmt *) * num_encodings,
>>> +                         GFP_KERNEL);
>>> +     if (!list)
>>> +             return -ENOMEM;
>>> +     node->supported_fmts.list = list;
>>> +
>>> +     for (i = 0, j = 0; i < num_encodings; i++) {
>>> +             const struct bcm2835_isp_fmt *fmt = get_fmt(fourccs[i]);
>>> +
>>> +             if (fmt) {
>>> +                     list[j] = fmt;
>>> +                     j++;
>>> +             }
>>> +     }
>>> +     node->supported_fmts.num_entries = j;
>>> +
>>> +     return 0;
>>> +}
>>> +
>>> +/*
>>> + * Register a device node /dev/video<N> to go along with one of the ISP's input
>>> + * or output nodes.
>>> + */
>>> +static int register_node(struct bcm2835_isp_dev *dev,
>>> +                      struct bcm2835_isp_node *node,
>>> +                      int index)
>>> +{
>>> +     struct video_device *vfd;
>>> +     struct vb2_queue *queue;
>>> +     int ret;
>>> +
>>> +     mutex_init(&node->lock);
>>> +
>>> +     node->dev = dev;
>>> +     vfd = &node->vfd;
>>> +     queue = &node->queue;
>>> +     queue->type = index_to_queue_type(index);
>>> +     /*
>>> +      * Setup the node type-specific params.
>>> +      *
>>> +      * Only the OUTPUT node can set controls and crop windows. However,
>>> +      * we must allow the s/g_selection ioctl on the stats node as v4l2
>>> +      * compliance expects it to return a -ENOTTY, and the framework
>>> +      * does not handle it if the ioctl is disabled.
>>> +      */
>>> +     switch (queue->type) {
>>> +     case V4L2_BUF_TYPE_VIDEO_OUTPUT:
>>> +             vfd->device_caps = V4L2_CAP_VIDEO_OUTPUT | V4L2_CAP_STREAMING;
>>> +             node->id = index;
>>> +             node->vfl_dir = VFL_DIR_TX;
>>> +             node->name = "output";
>>> +             break;
>>> +     case V4L2_BUF_TYPE_VIDEO_CAPTURE:
>>> +             vfd->device_caps = V4L2_CAP_VIDEO_CAPTURE | V4L2_CAP_STREAMING;
>>> +             /* First Capture node starts at id 0, etc. */
>>> +             node->id = index - BCM2835_ISP_NUM_OUTPUTS;
>>> +             node->vfl_dir = VFL_DIR_RX;
>>> +             node->name = "capture";
>>> +             v4l2_disable_ioctl(&node->vfd, VIDIOC_S_CTRL);
>>> +             v4l2_disable_ioctl(&node->vfd, VIDIOC_S_SELECTION);
>>> +             v4l2_disable_ioctl(&node->vfd, VIDIOC_G_SELECTION);
>>> +             break;
>>> +     case V4L2_BUF_TYPE_META_CAPTURE:
>>> +             vfd->device_caps = V4L2_CAP_META_CAPTURE | V4L2_CAP_STREAMING;
>>> +             node->id = index - BCM2835_ISP_NUM_OUTPUTS;
>>> +             node->vfl_dir = VFL_DIR_RX;
>>> +             node->name = "stats";
>>> +             v4l2_disable_ioctl(&node->vfd, VIDIOC_S_CTRL);
>>
>> Why not disable S/G_SELECTION for meta capture here rather than test for it
>> in the op functions?
> 
> I'd have to ask Naush. There were some very odd interactions with
> v4l2-compliance that he was struggling with.
> 
>>> +             break;
>>> +     }
>>> +
>>> +     /* We use the selection API instead of the old crop API. */
>>> +     v4l2_disable_ioctl(vfd, VIDIOC_CROPCAP);
>>> +     v4l2_disable_ioctl(vfd, VIDIOC_G_CROP);
>>> +     v4l2_disable_ioctl(vfd, VIDIOC_S_CROP);
>>
>> No need for this: the core handles this and will disable these ioctls
>> automatically.
> 
> The core will ENABLE these if g_selection and s_selection are defined,
> and uses mapping functions to try and convert between the two APIs.
> This may be down to missing the target check in s_selection, but again
> I seem to recall v4l2-compliance was reporting failures on these
> ioctls.

I suspect that the missing target check in g/s_selection is the cause of
the v4l2-compliance failures.

If there are still compliance issues after adding that check, then just
reach out to me for help.

Regards,

	Hans

  reply	other threads:[~2020-05-18 15:07 UTC|newest]

Thread overview: 104+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-05-04  9:25 [PATCH v2 00/34] Drivers for the BCM283x CSI-2/CCP2 receiver and ISP Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 01/34] media: uapi: v4l2-core: Add sensor ancillary data V4L2 fourcc type Laurent Pinchart
2020-05-04 13:48   ` Hans Verkuil
2020-05-04 14:39     ` Dave Stevenson
2020-05-04 15:32       ` Hans Verkuil
2020-05-04 16:08         ` Laurent Pinchart
2020-05-05 11:20           ` Dave Stevenson
2020-05-04  9:25 ` [PATCH v2 02/34] media: uapi: Add MEDIA_BUS_FMT_SENSOR_DATA media bus format Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 03/34] dt-bindings: media: Document BCM283x CSI2/CCP2 receiver Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 04/34] media: bcm2835-unicam: Driver for CCP2/CSI2 camera interface Laurent Pinchart
2020-05-04 15:21   ` Hans Verkuil
2020-05-05  1:26   ` kbuild test robot
2020-05-06 18:01   ` Nicolas Saenz Julienne
2020-08-29 11:20   ` Jacopo Mondi
2020-08-29 18:32     ` Laurent Pinchart
2020-08-31  7:38       ` Jacopo Mondi
2020-08-31 14:17         ` Laurent Pinchart
2020-08-31 14:46           ` Jacopo Mondi
2020-08-31 14:56             ` Laurent Pinchart
2020-09-01  8:41               ` Dave Stevenson
2020-09-01 10:22                 ` Jacopo Mondi
2020-09-01 16:37                   ` Dave Stevenson
2020-09-01 17:11                     ` Laurent Pinchart
2020-09-15  7:03   ` Sakari Ailus
2020-09-15  9:32     ` Laurent Pinchart
2020-09-15 13:28       ` Dave Stevenson
2020-10-30 17:53         ` Jacopo Mondi
2020-09-15 17:30     ` Dave Stevenson
2020-05-04  9:25 ` [PATCH v2 05/34] ARM: dts: bcm2711: Add Unicam DT nodes Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 06/34] staging: vc04_services: Add new vc-sm-cma driver Laurent Pinchart
2020-05-05  2:38   ` kbuild test robot
2020-05-05 12:17   ` kbuild test robot
2020-05-06  3:05   ` kbuild test robot
2020-05-06 18:04   ` Nicolas Saenz Julienne
2020-05-06 19:24     ` Dave Stevenson
2020-05-08  0:11       ` Laurent Pinchart
2020-05-18 12:06         ` Hans Verkuil
2020-08-24 16:39       ` Jacopo Mondi
2020-08-25 17:52         ` Dave Stevenson
2020-08-27 10:38           ` Jacopo Mondi
2020-08-27 12:51             ` Dave Stevenson
2020-08-27 16:46               ` Jacopo Mondi
2020-08-27 17:19                 ` Dave Stevenson
2020-05-11 18:42   ` Nicolas Saenz Julienne
2020-05-18 15:48     ` Dave Stevenson
2020-05-20 14:41       ` Nicolas Saenz Julienne
2020-05-21 11:01         ` Dave Stevenson
2020-05-04  9:25 ` [PATCH v2 07/34] staging: bcm2835: Break MMAL support out from camera Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 08/34] staging: mmal-vchiq: Allocate and free components as required Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 09/34] staging: mmal-vchiq: Avoid use of bool in structures Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 10/34] staging: mmal-vchiq: Make timeout a defined parameter Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 11/34] staging: mmal-vchiq: Make a mmal_buf struct for passing parameters Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 12/34] staging: mmal-vchiq: Add support for event callbacks Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 13/34] staging: mmal-vchiq: Support sending data to MMAL ports Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 14/34] staging: mmal-vchiq: Fixup vchiq-mmal include ordering Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 15/34] staging: mmal-vchiq: Use vc-sm-cma to support zero copy Laurent Pinchart
2020-05-04 16:30   ` Nicolas Saenz Julienne
2020-05-05 16:07   ` kbuild test robot
2020-05-11 19:15   ` Nicolas Saenz Julienne
2020-05-04  9:25 ` [PATCH v2 16/34] staging: mmal-vchiq: Fix client_component for 64 bit kernel Laurent Pinchart
2020-05-18 10:19   ` Hans Verkuil
2020-05-04  9:25 ` [PATCH v2 17/34] staging: mmal_vchiq: Add in the Bayer encoding formats Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 18/34] staging: mmal-vchiq: Always return the param size from param_get Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 19/34] staging: mmal-vchiq: If the VPU returns an error, don't negate it Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 20/34] staging: mmal-vchiq: Fix handling of VB2_MEMORY_DMABUF buffers Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 21/34] staging: mmal-vchiq: Update mmal_parameters.h with recently defined params Laurent Pinchart
2020-05-04  9:25 ` [PATCH v2 22/34] staging: mmal-vchiq: Free the event context for control ports Laurent Pinchart
2020-05-04  9:26 ` [PATCH v2 23/34] staging: mmal-vchiq: Fix memory leak in error path Laurent Pinchart
2020-05-04  9:26 ` [PATCH v2 24/34] staging: mmal-vchiq: Fix formatting errors in mmal_parameters.h Laurent Pinchart
2020-05-04  9:26 ` [PATCH v2 25/34] staging: vchiq_arm: Register vcsm-cma as a platform driver Laurent Pinchart
2020-05-04  9:26 ` [PATCH v2 26/34] staging: vchiq_arm: Set up dma ranges on child devices Laurent Pinchart
2020-05-04 16:54   ` Nicolas Saenz Julienne
2020-08-25 16:57     ` Jacopo Mondi
2020-05-04  9:26 ` [PATCH v2 27/34] staging: vchiq: Use the old dma controller for OF config on platform devices Laurent Pinchart
2020-05-04 15:44   ` Nicolas Saenz Julienne
2020-05-04  9:26 ` [PATCH v2 28/34] staging: vchiq_2835_arm: Implement a DMA pool for small bulk transfers Laurent Pinchart
2020-05-04  9:26 ` [PATCH v2 29/34] staging: vchiq: Add 36-bit address support Laurent Pinchart
2020-05-04 17:40   ` Nicolas Saenz Julienne
2020-05-04 20:46     ` Phil Elwell
2020-05-05 10:13       ` Nicolas Saenz Julienne
2020-05-05 10:57         ` Phil Elwell
2020-05-05 13:22   ` kbuild test robot
2020-05-04  9:26 ` [PATCH v2 30/34] staging: vchiq_arm: Give vchiq children DT nodes Laurent Pinchart
2020-05-04 17:12   ` Nicolas Saenz Julienne
2020-05-04 19:42     ` Phil Elwell
2020-05-05 10:37       ` Nicolas Saenz Julienne
2020-05-05 10:50         ` Phil Elwell
2020-05-04  9:26 ` [PATCH v2 31/34] staging: vchiq_arm: Add a matching unregister call Laurent Pinchart
2020-05-04  9:26 ` [PATCH v2 32/34] media: videobuf2: Allow exporting of a struct dmabuf Laurent Pinchart
2020-05-04 13:36   ` Hans Verkuil
2020-05-04  9:26 ` [PATCH v2 33/34] staging: bcm2835-isp: Add support for BC2835 ISP Laurent Pinchart
2020-05-11 19:19   ` Nicolas Saenz Julienne
2020-05-18 13:38     ` Dave Stevenson
2020-05-20 13:46       ` Nicolas Saenz Julienne
2020-05-18 12:02   ` Hans Verkuil
2020-05-18 14:36     ` Dave Stevenson
2020-05-18 15:07       ` Hans Verkuil [this message]
2020-05-19 12:47     ` Naushir Patuck
2020-05-19 13:03       ` Jacopo Mondi
2020-06-24 11:28       ` Hans Verkuil
2020-05-04  9:26 ` [PATCH v2 34/34] staging: vchiq: Load bcm2835_isp driver from vchiq Laurent Pinchart
2020-05-04 15:15 ` [PATCH v2 00/34] Drivers for the BCM283x CSI-2/CCP2 receiver and ISP Nicolas Saenz Julienne
2020-05-04 15:38   ` Laurent Pinchart
2020-05-04 16:15     ` Nicolas Saenz Julienne

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=bdcb471b-e066-373f-aa16-b7caf5c52795@xs4all.nl \
    --to=hverkuil@xs4all.nl \
    --cc=dave.stevenson@raspberrypi.com \
    --cc=jacopo@jmondi.org \
    --cc=kieran.bingham@ideasonboard.com \
    --cc=laurent.pinchart@ideasonboard.com \
    --cc=linux-media@vger.kernel.org \
    --cc=naush@raspberrypi.com \
    --cc=niklas.soderlund@ragnatech.se \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).