kernfs: implement kernfs_ops->atomic_write_len
authorTejun Heo <tj@kernel.org>
Mon, 3 Feb 2014 19:09:13 +0000 (14:09 -0500)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Fri, 7 Feb 2014 23:52:48 +0000 (15:52 -0800)
commit4d3773c4bb41ed5228f1ab7a4a52b79e17b10515
tree71c30d7009b36d22e9aab816f3dc7e373bb5f1e9
parentd35258ef702cca0c4e66d799f8e38b78c02ce8a5
kernfs: implement kernfs_ops->atomic_write_len

A write to a kernfs_node is buffered through a kernel buffer.  Writes
<= PAGE_SIZE are performed atomically, while larger ones are executed
in PAGE_SIZE chunks.  While this is enough for sysfs, cgroup which is
scheduled to be converted to use kernfs needs a bit more control over
it.

This patch adds kernfs_ops->atomic_write_len.  If not set (zero), the
behavior stays the same.  If set, writes upto the size are executed
atomically and larger writes are rejected with -E2BIG.

A different implementation strategy would be allowing configuring
chunking size while making the original write size available to the
write method; however, such strategy, while being more complicated,
doesn't really buy anything.  If the write implementation has to
handle chunking, the specific chunk size shouldn't matter all that
much.

Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/kernfs/file.c
include/linux/kernfs.h