linux/Documentation/DocBook/media/v4l/vidioc-create-bufs.xml

167 lines
6.1 KiB
XML
Raw Normal View History

<refentry id="vidioc-create-bufs">
<refmeta>
<refentrytitle>ioctl VIDIOC_CREATE_BUFS</refentrytitle>
&manvol;
</refmeta>
<refnamediv>
<refname>VIDIOC_CREATE_BUFS</refname>
<refpurpose>Create buffers for Memory Mapped or User Pointer or DMA Buffer
I/O</refpurpose>
</refnamediv>
<refsynopsisdiv>
<funcsynopsis>
<funcprototype>
<funcdef>int <function>ioctl</function></funcdef>
<paramdef>int <parameter>fd</parameter></paramdef>
<paramdef>int <parameter>request</parameter></paramdef>
<paramdef>struct v4l2_create_buffers *<parameter>argp</parameter></paramdef>
</funcprototype>
</funcsynopsis>
</refsynopsisdiv>
<refsect1>
<title>Arguments</title>
<variablelist>
<varlistentry>
<term><parameter>fd</parameter></term>
<listitem>
<para>&fd;</para>
</listitem>
</varlistentry>
<varlistentry>
<term><parameter>request</parameter></term>
<listitem>
<para>VIDIOC_CREATE_BUFS</para>
</listitem>
</varlistentry>
<varlistentry>
<term><parameter>argp</parameter></term>
<listitem>
<para></para>
</listitem>
</varlistentry>
</variablelist>
</refsect1>
<refsect1>
<title>Description</title>
<note>
<title>Experimental</title>
<para>This is an <link linkend="experimental"> experimental </link>
interface and may change in the future.</para>
</note>
<para>This ioctl is used to create buffers for <link linkend="mmap">memory
mapped</link> or <link linkend="userp">user pointer</link> or <link
linkend="dmabuf">DMA buffer</link> I/O. It can be used as an alternative or in
addition to the <constant>VIDIOC_REQBUFS</constant> ioctl, when a tighter
control over buffers is required. This ioctl can be called multiple times to
create buffers of different sizes.</para>
<para>To allocate the device buffers applications must initialize the
relevant fields of the <structname>v4l2_create_buffers</structname> structure.
The <structfield>count</structfield> field must be set to the number of
requested buffers, the <structfield>memory</structfield> field specifies the
requested I/O method and the <structfield>reserved</structfield> array must be
zeroed.</para>
<para>The <structfield>format</structfield> field specifies the image format
that the buffers must be able to handle. The application has to fill in this
&v4l2-format;. Usually this will be done using the
<constant>VIDIOC_TRY_FMT</constant> or <constant>VIDIOC_G_FMT</constant> ioctl()
to ensure that the requested format is supported by the driver. Unsupported
formats will result in an error.</para>
<para>The buffers created by this ioctl will have as minimum size the size
defined by the <structfield>format.pix.sizeimage</structfield> field. If the
<structfield>format.pix.sizeimage</structfield> field is less than the minimum
required for the given format, then <structfield>sizeimage</structfield> will be
increased by the driver to that minimum to allocate the buffers. If it is
larger, then the value will be used as-is. The same applies to the
<structfield>sizeimage</structfield> field of the
<structname>v4l2_plane_pix_format</structname> structure in the case of
multiplanar formats.</para>
<para>When the ioctl is called with a pointer to this structure the driver
will attempt to allocate up to the requested number of buffers and store the
actual number allocated and the starting index in the
<structfield>count</structfield> and the <structfield>index</structfield> fields
respectively. On return <structfield>count</structfield> can be smaller than
the number requested. The driver may also increase buffer sizes if required,
however, it will not update <structfield>sizeimage</structfield> field values.
The user has to use <constant>VIDIOC_QUERYBUF</constant> to retrieve that
information.</para>
<table pgwide="1" frame="none" id="v4l2-create-buffers">
<title>struct <structname>v4l2_create_buffers</structname></title>
<tgroup cols="3">
&cs-str;
<tbody valign="top">
<row>
<entry>__u32</entry>
<entry><structfield>index</structfield></entry>
<entry>The starting buffer index, returned by the driver.</entry>
</row>
<row>
<entry>__u32</entry>
<entry><structfield>count</structfield></entry>
<entry>The number of buffers requested or granted. If count == 0, then
<constant>VIDIOC_CREATE_BUFS</constant> will set <structfield>index</structfield>
to the current number of created buffers, and it will check the validity of
<structfield>memory</structfield> and <structfield>format.type</structfield>.
If those are invalid -1 is returned and errno is set to &EINVAL;,
otherwise <constant>VIDIOC_CREATE_BUFS</constant> returns 0. It will
never set errno to &EBUSY; in this particular case.</entry>
</row>
<row>
[media] v4l2: use __u32 rather than enums in ioctl() structs V4L2 uses the enum type in IOCTL arguments in IOCTLs that were defined until the use of enum was considered less than ideal. Recently Rémi Denis-Courmont brought up the issue by proposing a patch to convert the enums to unsigned: <URL:http://www.spinics.net/lists/linux-media/msg46167.html> This sparked a long discussion where another solution to the issue was proposed: two sets of IOCTL structures, one with __u32 and the other with enums, and conversion code between the two: <URL:http://www.spinics.net/lists/linux-media/msg47168.html> Both approaches implement a complete solution that resolves the problem. The first one is simple but requires assuming enums and __u32 are the same in size (so we won't break the ABI) while the second one is more complex and less clean but does not require making that assumption. The issue boils down to whether enums are fundamentally different from __u32 or not, and can the former be substituted by the latter. During the discussion it was concluded that the __u32 has the same size as enums on all archs Linux is supported: it has not been shown that replacing those enums in IOCTL arguments would break neither source or binary compatibility. If no such reason is found, just replacing the enums with __u32s is the way to go. This is what this patch does. This patch is slightly different from Remi's first RFC (link above): it uses __u32 instead of unsigned and also changes the arguments of VIDIOC_G_PRIORITY and VIDIOC_S_PRIORITY. Signed-off-by: Rémi Denis-Courmont <remi@remlab.net> Signed-off-by: Sakari Ailus <sakari.ailus@iki.fi> Acked-by: Hans Verkuil <hans.verkuil@cisco.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2012-05-10 13:02:07 +08:00
<entry>__u32</entry>
<entry><structfield>memory</structfield></entry>
<entry>Applications set this field to
<constant>V4L2_MEMORY_MMAP</constant>,
<constant>V4L2_MEMORY_DMABUF</constant> or
[media] v4l2: use __u32 rather than enums in ioctl() structs V4L2 uses the enum type in IOCTL arguments in IOCTLs that were defined until the use of enum was considered less than ideal. Recently Rémi Denis-Courmont brought up the issue by proposing a patch to convert the enums to unsigned: <URL:http://www.spinics.net/lists/linux-media/msg46167.html> This sparked a long discussion where another solution to the issue was proposed: two sets of IOCTL structures, one with __u32 and the other with enums, and conversion code between the two: <URL:http://www.spinics.net/lists/linux-media/msg47168.html> Both approaches implement a complete solution that resolves the problem. The first one is simple but requires assuming enums and __u32 are the same in size (so we won't break the ABI) while the second one is more complex and less clean but does not require making that assumption. The issue boils down to whether enums are fundamentally different from __u32 or not, and can the former be substituted by the latter. During the discussion it was concluded that the __u32 has the same size as enums on all archs Linux is supported: it has not been shown that replacing those enums in IOCTL arguments would break neither source or binary compatibility. If no such reason is found, just replacing the enums with __u32s is the way to go. This is what this patch does. This patch is slightly different from Remi's first RFC (link above): it uses __u32 instead of unsigned and also changes the arguments of VIDIOC_G_PRIORITY and VIDIOC_S_PRIORITY. Signed-off-by: Rémi Denis-Courmont <remi@remlab.net> Signed-off-by: Sakari Ailus <sakari.ailus@iki.fi> Acked-by: Hans Verkuil <hans.verkuil@cisco.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
2012-05-10 13:02:07 +08:00
<constant>V4L2_MEMORY_USERPTR</constant>. See <xref linkend="v4l2-memory"
/></entry>
</row>
<row>
<entry>&v4l2-format;</entry>
<entry><structfield>format</structfield></entry>
<entry>Filled in by the application, preserved by the driver.</entry>
</row>
<row>
<entry>__u32</entry>
<entry><structfield>reserved</structfield>[8]</entry>
<entry>A place holder for future extensions. Drivers and applications
must set the array to zero.</entry>
</row>
</tbody>
</tgroup>
</table>
</refsect1>
<refsect1>
&return-value;
<variablelist>
<varlistentry>
<term><errorcode>ENOMEM</errorcode></term>
<listitem>
<para>No memory to allocate buffers for <link linkend="mmap">memory
mapped</link> I/O.</para>
</listitem>
</varlistentry>
<varlistentry>
<term><errorcode>EINVAL</errorcode></term>
<listitem>
<para>The buffer type (<structfield>format.type</structfield> field),
requested I/O method (<structfield>memory</structfield>) or format
(<structfield>format</structfield> field) is not valid.</para>
</listitem>
</varlistentry>
</variablelist>
</refsect1>
</refentry>