[lttng-modules.git] / lib / ringbuffer / ring_buffer_splice.c

/* SPDX-License-Identifier: GPL-2.0
 *
 * ring_buffer_splice.c
 *
 * Copyright (C) 2002-2005 - Tom Zanussi <zanussi@us.ibm.com>, IBM Corp
 * Copyright (C) 1999-2005 - Karim Yaghmour <karim@opersys.com>
 * Copyright (C) 2008-2012 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
 *
 * Re-using code from kernel/relay.c, which is why it is licensed under
 * the GPL-2.0.
 */

#include <linux/module.h>
#include <linux/fs.h>
#include <linux/version.h>

#include <wrapper/splice.h>
#include <wrapper/ringbuffer/backend.h>
#include <wrapper/ringbuffer/frontend.h>
#include <wrapper/ringbuffer/vfs.h>

#if 0
#define printk_dbg(fmt, args...) printk(fmt, args)
#else
#define printk_dbg(fmt, args...)
#endif

loff_t vfs_lib_ring_buffer_no_llseek(struct file *file, loff_t offset,
		int origin)
{
	return -ESPIPE;
}
EXPORT_SYMBOL_GPL(vfs_lib_ring_buffer_no_llseek);

/*
 * Release pages from the buffer so splice pipe_to_file can move them.
 * Called after the pipe has been populated with buffer pages.
 */
static void lib_ring_buffer_pipe_buf_release(struct pipe_inode_info *pipe,
					     struct pipe_buffer *pbuf)
{
	__free_page(pbuf->page);
}

static const struct pipe_buf_operations ring_buffer_pipe_buf_ops = {
	.can_merge = 0,
#if (LINUX_VERSION_CODE < KERNEL_VERSION(3,15,0))
	.map = generic_pipe_buf_map,
	.unmap = generic_pipe_buf_unmap,
#endif
	.confirm = generic_pipe_buf_confirm,
	.release = lib_ring_buffer_pipe_buf_release,
	.steal = generic_pipe_buf_steal,
	.get = generic_pipe_buf_get,
};

/*
 * Page release operation after splice pipe_to_file ends.
 */
static void lib_ring_buffer_page_release(struct splice_pipe_desc *spd,
					 unsigned int i)
{
	__free_page(spd->pages[i]);
}

/*
 *	subbuf_splice_actor - splice up to one subbuf's worth of data
 */
static int subbuf_splice_actor(struct file *in,
			       loff_t *ppos,
			       struct pipe_inode_info *pipe,
			       size_t len,
			       unsigned int flags,
			       struct lib_ring_buffer *buf)
{
	struct channel *chan = buf->backend.chan;
	const struct lib_ring_buffer_config *config = &chan->backend.config;
	unsigned int poff, subbuf_pages, nr_pages;
	struct page *pages[PIPE_DEF_BUFFERS];
	struct partial_page partial[PIPE_DEF_BUFFERS];
	struct splice_pipe_desc spd = {
		.pages = pages,
		.nr_pages = 0,
		.partial = partial,
#if (LINUX_VERSION_CODE < KERNEL_VERSION(4,12,0))
		.flags = flags,
#endif
		.ops = &ring_buffer_pipe_buf_ops,
		.spd_release = lib_ring_buffer_page_release,
	};
	unsigned long consumed_old, roffset;
	unsigned long bytes_avail;

	/*
	 * Check that a GET_SUBBUF ioctl has been done before.
	 */
	WARN_ON(atomic_long_read(&buf->active_readers) != 1);
	consumed_old = lib_ring_buffer_get_consumed(config, buf);
	consumed_old += *ppos;

	/*
	 * Adjust read len, if longer than what is available.
	 * Max read size is 1 subbuffer due to get_subbuf/put_subbuf for
	 * protection.
	 */
	bytes_avail = chan->backend.subbuf_size;
	WARN_ON(bytes_avail > chan->backend.buf_size);
	len = min_t(size_t, len, bytes_avail);
	subbuf_pages = bytes_avail >> PAGE_SHIFT;
	nr_pages = min_t(unsigned int, subbuf_pages, PIPE_DEF_BUFFERS);
	roffset = consumed_old & PAGE_MASK;
	poff = consumed_old & ~PAGE_MASK;
	printk_dbg(KERN_DEBUG "SPLICE actor len %zu pos %zd write_pos %ld\n",
		   len, (ssize_t)*ppos, lib_ring_buffer_get_offset(config, buf));

	for (; spd.nr_pages < nr_pages; spd.nr_pages++) {
		unsigned int this_len;
		unsigned long *pfnp, new_pfn;
		struct page *new_page;
		void **virt;

		if (!len)
			break;
		printk_dbg(KERN_DEBUG "SPLICE actor loop len %zu roffset %ld\n",
			   len, roffset);

		/*
		 * We have to replace the page we are moving into the splice
		 * pipe.
		 */
		new_page = alloc_pages_node(cpu_to_node(max(buf->backend.cpu,
							    0)),
					    GFP_KERNEL | __GFP_ZERO, 0);
		if (!new_page)
			break;
		new_pfn = page_to_pfn(new_page);
		this_len = PAGE_SIZE - poff;
		pfnp = lib_ring_buffer_read_get_pfn(&buf->backend, roffset, &virt);
		spd.pages[spd.nr_pages] = pfn_to_page(*pfnp);
		*pfnp = new_pfn;
		*virt = page_address(new_page);
		spd.partial[spd.nr_pages].offset = poff;
		spd.partial[spd.nr_pages].len = this_len;

		poff = 0;
		roffset += PAGE_SIZE;
		len -= this_len;
	}

	if (!spd.nr_pages)
		return 0;

	return wrapper_splice_to_pipe(pipe, &spd);
}

ssize_t lib_ring_buffer_splice_read(struct file *in, loff_t *ppos,
				    struct pipe_inode_info *pipe, size_t len,
				    unsigned int flags,
				    struct lib_ring_buffer *buf)
{
	struct channel *chan = buf->backend.chan;
	const struct lib_ring_buffer_config *config = &chan->backend.config;
	ssize_t spliced;
	int ret;

	if (config->output != RING_BUFFER_SPLICE)
		return -EINVAL;

	/*
	 * We require ppos and length to be page-aligned for performance reasons
	 * (no page copy). Size is known using the ioctl
	 * RING_BUFFER_GET_PADDED_SUBBUF_SIZE, which is page-size padded.
	 * We fail when the ppos or len passed is not page-sized, because splice
	 * is not allowed to copy more than the length passed as parameter (so
	 * the ABI does not let us silently copy more than requested to include
	 * padding).
	 */
	if (*ppos != PAGE_ALIGN(*ppos) || len != PAGE_ALIGN(len))
		return -EINVAL;

	ret = 0;
	spliced = 0;

	printk_dbg(KERN_DEBUG "SPLICE read len %zu pos %zd\n", len,
		   (ssize_t)*ppos);
	while (len && !spliced) {
		ret = subbuf_splice_actor(in, ppos, pipe, len, flags, buf);
		printk_dbg(KERN_DEBUG "SPLICE read loop ret %d\n", ret);
		if (ret < 0)
			break;
		else if (!ret) {
			if (flags & SPLICE_F_NONBLOCK)
				ret = -EAGAIN;
			break;
		}

		*ppos += ret;
		if (ret > len)
			len = 0;
		else
			len -= ret;
		spliced += ret;
	}

	if (spliced)
		return spliced;

	return ret;
}
EXPORT_SYMBOL_GPL(lib_ring_buffer_splice_read);

ssize_t vfs_lib_ring_buffer_splice_read(struct file *in, loff_t *ppos,
				    struct pipe_inode_info *pipe, size_t len,
				    unsigned int flags)
{
	struct lib_ring_buffer *buf = in->private_data;

	return lib_ring_buffer_splice_read(in, ppos, pipe, len, flags, buf);
}
EXPORT_SYMBOL_GPL(vfs_lib_ring_buffer_splice_read);
Commit	Line	Data
9f36eaed MJ	1	/* SPDX-License-Identifier: GPL-2.0
9f36eaed MJ	2	*
f3bc08c5 MD	3	* ring_buffer_splice.c
	4	*
	5	* Copyright (C) 2002-2005 - Tom Zanussi <zanussi@us.ibm.com>, IBM Corp
	6	* Copyright (C) 1999-2005 - Karim Yaghmour <karim@opersys.com>
886d51a3	7	* Copyright (C) 2008-2012 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
f3bc08c5	8	*
886d51a3	9	* Re-using code from kernel/relay.c, which is why it is licensed under
9f36eaed	10	* the GPL-2.0.
f3bc08c5 MD	11	*/
	12
	13	#include <linux/module.h>
	14	#include <linux/fs.h>
989f58e8	15	#include <linux/version.h>
f3bc08c5	16
c075712b MD	17	#include <wrapper/splice.h>
	18	#include <wrapper/ringbuffer/backend.h>
	19	#include <wrapper/ringbuffer/frontend.h>
	20	#include <wrapper/ringbuffer/vfs.h>
f3bc08c5 MD	21
	22	#if 0
	23	#define printk_dbg(fmt, args...) printk(fmt, args)
	24	#else
	25	#define printk_dbg(fmt, args...)
	26	#endif
	27
d83004aa JD	28	loff_t vfs_lib_ring_buffer_no_llseek(struct file *file, loff_t offset,
d83004aa JD	29	int origin)
f3bc08c5 MD	30	{
	31	return -ESPIPE;
	32	}
d83004aa	33	EXPORT_SYMBOL_GPL(vfs_lib_ring_buffer_no_llseek);
f3bc08c5 MD	34
	35	/*
	36	* Release pages from the buffer so splice pipe_to_file can move them.
	37	* Called after the pipe has been populated with buffer pages.
	38	*/
	39	static void lib_ring_buffer_pipe_buf_release(struct pipe_inode_info *pipe,
	40	struct pipe_buffer *pbuf)
	41	{
	42	__free_page(pbuf->page);
	43	}
	44
	45	static const struct pipe_buf_operations ring_buffer_pipe_buf_ops = {
	46	.can_merge = 0,
989f58e8	47	#if (LINUX_VERSION_CODE < KERNEL_VERSION(3,15,0))
f3bc08c5 MD	48	.map = generic_pipe_buf_map,
f3bc08c5 MD	49	.unmap = generic_pipe_buf_unmap,
989f58e8	50	#endif
f3bc08c5 MD	51	.confirm = generic_pipe_buf_confirm,
	52	.release = lib_ring_buffer_pipe_buf_release,
	53	.steal = generic_pipe_buf_steal,
	54	.get = generic_pipe_buf_get,
	55	};
	56
	57	/*
	58	* Page release operation after splice pipe_to_file ends.
	59	*/
	60	static void lib_ring_buffer_page_release(struct splice_pipe_desc *spd,
	61	unsigned int i)
	62	{
	63	__free_page(spd->pages[i]);
	64	}
	65
	66	/*
	67	* subbuf_splice_actor - splice up to one subbuf's worth of data
	68	*/
	69	static int subbuf_splice_actor(struct file *in,
	70	loff_t *ppos,
	71	struct pipe_inode_info *pipe,
	72	size_t len,
d83004aa JD	73	unsigned int flags,
d83004aa JD	74	struct lib_ring_buffer *buf)
f3bc08c5	75	{
f3bc08c5	76	struct channel *chan = buf->backend.chan;
5a8fd222	77	const struct lib_ring_buffer_config *config = &chan->backend.config;
f3bc08c5 MD	78	unsigned int poff, subbuf_pages, nr_pages;
	79	struct page *pages[PIPE_DEF_BUFFERS];
	80	struct partial_page partial[PIPE_DEF_BUFFERS];
	81	struct splice_pipe_desc spd = {
	82	.pages = pages,
	83	.nr_pages = 0,
	84	.partial = partial,
e2833bed	85	#if (LINUX_VERSION_CODE < KERNEL_VERSION(4,12,0))
f3bc08c5	86	.flags = flags,
e2833bed	87	#endif
f3bc08c5 MD	88	.ops = &ring_buffer_pipe_buf_ops,
	89	.spd_release = lib_ring_buffer_page_release,
	90	};
88dfd899	91	unsigned long consumed_old, roffset;
f3bc08c5 MD	92	unsigned long bytes_avail;
	93
	94	/*
	95	* Check that a GET_SUBBUF ioctl has been done before.
	96	*/
	97	WARN_ON(atomic_long_read(&buf->active_readers) != 1);
	98	consumed_old = lib_ring_buffer_get_consumed(config, buf);
	99	consumed_old += *ppos;
f3bc08c5 MD	100
	101	/*
	102	* Adjust read len, if longer than what is available.
	103	* Max read size is 1 subbuffer due to get_subbuf/put_subbuf for
	104	* protection.
	105	*/
	106	bytes_avail = chan->backend.subbuf_size;
	107	WARN_ON(bytes_avail > chan->backend.buf_size);
	108	len = min_t(size_t, len, bytes_avail);
	109	subbuf_pages = bytes_avail >> PAGE_SHIFT;
	110	nr_pages = min_t(unsigned int, subbuf_pages, PIPE_DEF_BUFFERS);
	111	roffset = consumed_old & PAGE_MASK;
	112	poff = consumed_old & ~PAGE_MASK;
	113	printk_dbg(KERN_DEBUG "SPLICE actor len %zu pos %zd write_pos %ld\n",
	114	len, (ssize_t)*ppos, lib_ring_buffer_get_offset(config, buf));
	115
	116	for (; spd.nr_pages < nr_pages; spd.nr_pages++) {
	117	unsigned int this_len;
0112cb7b MD	118	unsigned long *pfnp, new_pfn;
0112cb7b MD	119	struct page *new_page;
f3bc08c5 MD	120	void **virt;
	121
	122	if (!len)
	123	break;
	124	printk_dbg(KERN_DEBUG "SPLICE actor loop len %zu roffset %ld\n",
	125	len, roffset);
	126
	127	/*
	128	* We have to replace the page we are moving into the splice
	129	* pipe.
	130	*/
	131	new_page = alloc_pages_node(cpu_to_node(max(buf->backend.cpu,
	132	0)),
	133	GFP_KERNEL \| __GFP_ZERO, 0);
	134	if (!new_page)
	135	break;
0112cb7b	136	new_pfn = page_to_pfn(new_page);
f3bc08c5	137	this_len = PAGE_SIZE - poff;
0112cb7b MD	138	pfnp = lib_ring_buffer_read_get_pfn(&buf->backend, roffset, &virt);
	139	spd.pages[spd.nr_pages] = pfn_to_page(*pfnp);
	140	*pfnp = new_pfn;
f3bc08c5 MD	141	*virt = page_address(new_page);
	142	spd.partial[spd.nr_pages].offset = poff;
	143	spd.partial[spd.nr_pages].len = this_len;
	144
	145	poff = 0;
	146	roffset += PAGE_SIZE;
	147	len -= this_len;
	148	}
	149
	150	if (!spd.nr_pages)
	151	return 0;
	152
90225db5	153	return wrapper_splice_to_pipe(pipe, &spd);
f3bc08c5 MD	154	}
	155
	156	ssize_t lib_ring_buffer_splice_read(struct file in, loff_t ppos,
	157	struct pipe_inode_info *pipe, size_t len,
d83004aa JD	158	unsigned int flags,
d83004aa JD	159	struct lib_ring_buffer *buf)
f3bc08c5	160	{
f3bc08c5	161	struct channel *chan = buf->backend.chan;
5a8fd222	162	const struct lib_ring_buffer_config *config = &chan->backend.config;
f3bc08c5 MD	163	ssize_t spliced;
	164	int ret;
	165
	166	if (config->output != RING_BUFFER_SPLICE)
	167	return -EINVAL;
	168
	169	/*
	170	* We require ppos and length to be page-aligned for performance reasons
	171	* (no page copy). Size is known using the ioctl
	172	* RING_BUFFER_GET_PADDED_SUBBUF_SIZE, which is page-size padded.
	173	* We fail when the ppos or len passed is not page-sized, because splice
	174	* is not allowed to copy more than the length passed as parameter (so
	175	* the ABI does not let us silently copy more than requested to include
	176	* padding).
	177	*/
	178	if (ppos != PAGE_ALIGN(ppos) \|\| len != PAGE_ALIGN(len))
	179	return -EINVAL;
	180
	181	ret = 0;
	182	spliced = 0;
	183
	184	printk_dbg(KERN_DEBUG "SPLICE read len %zu pos %zd\n", len,
	185	(ssize_t)*ppos);
	186	while (len && !spliced) {
d83004aa	187	ret = subbuf_splice_actor(in, ppos, pipe, len, flags, buf);
f3bc08c5 MD	188	printk_dbg(KERN_DEBUG "SPLICE read loop ret %d\n", ret);
	189	if (ret < 0)
	190	break;
	191	else if (!ret) {
	192	if (flags & SPLICE_F_NONBLOCK)
	193	ret = -EAGAIN;
	194	break;
	195	}
	196
	197	*ppos += ret;
	198	if (ret > len)
	199	len = 0;
	200	else
	201	len -= ret;
	202	spliced += ret;
	203	}
	204
	205	if (spliced)
	206	return spliced;
	207
	208	return ret;
	209	}
	210	EXPORT_SYMBOL_GPL(lib_ring_buffer_splice_read);
d83004aa JD	211
	212	ssize_t vfs_lib_ring_buffer_splice_read(struct file in, loff_t ppos,
	213	struct pipe_inode_info *pipe, size_t len,
	214	unsigned int flags)
	215	{
	216	struct lib_ring_buffer *buf = in->private_data;
	217
	218	return lib_ring_buffer_splice_read(in, ppos, pipe, len, flags, buf);
	219	}
	220	EXPORT_SYMBOL_GPL(vfs_lib_ring_buffer_splice_read);