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

/* SPDX-License-Identifier: GPL-2.0-only
 *
 * 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 <ringbuffer/backend.h>
#include <ringbuffer/frontend.h>
#include <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);
}

#if (LINUX_VERSION_CODE >= KERNEL_VERSION(5,8,0))
static const struct pipe_buf_operations ring_buffer_pipe_buf_ops = {
	.release = lib_ring_buffer_pipe_buf_release,
	.try_steal = generic_pipe_buf_try_steal,
	.get = generic_pipe_buf_get
};
#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(5,1,0))
static const struct pipe_buf_operations ring_buffer_pipe_buf_ops = {
	.confirm = generic_pipe_buf_confirm,
	.release = lib_ring_buffer_pipe_buf_release,
	.steal = generic_pipe_buf_steal,
	.get = generic_pipe_buf_get
};
#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(3,15,0))
static const struct pipe_buf_operations ring_buffer_pipe_buf_ops = {
	.can_merge = 0,
	.confirm = generic_pipe_buf_confirm,
	.release = lib_ring_buffer_pipe_buf_release,
	.steal = generic_pipe_buf_steal,
	.get = generic_pipe_buf_get
};
#else
static const struct pipe_buf_operations ring_buffer_pipe_buf_ops = {
	.can_merge = 0,
	.map = generic_pipe_buf_map,
	.unmap = generic_pipe_buf_unmap,
	.confirm = generic_pipe_buf_confirm,
	.release = lib_ring_buffer_pipe_buf_release,
	.steal = generic_pipe_buf_steal,
	.get = generic_pipe_buf_get
};
#endif

/*
 * 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
b7cdc182	1	/* SPDX-License-Identifier: GPL-2.0-only
9f36eaed	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	17	#include <wrapper/splice.h>
24591303 MD	18	#include <ringbuffer/backend.h>
	19	#include <ringbuffer/frontend.h>
	20	#include <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
4352d508 MJ	45	#if (LINUX_VERSION_CODE >= KERNEL_VERSION(5,8,0))
	46	static const struct pipe_buf_operations ring_buffer_pipe_buf_ops = {
	47	.release = lib_ring_buffer_pipe_buf_release,
	48	.try_steal = generic_pipe_buf_try_steal,
	49	.get = generic_pipe_buf_get
	50	};
	51	#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(5,1,0))
	52	static const struct pipe_buf_operations ring_buffer_pipe_buf_ops = {
	53	.confirm = generic_pipe_buf_confirm,
	54	.release = lib_ring_buffer_pipe_buf_release,
	55	.steal = generic_pipe_buf_steal,
	56	.get = generic_pipe_buf_get
	57	};
	58	#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(3,15,0))
	59	static const struct pipe_buf_operations ring_buffer_pipe_buf_ops = {
	60	.can_merge = 0,
	61	.confirm = generic_pipe_buf_confirm,
	62	.release = lib_ring_buffer_pipe_buf_release,
	63	.steal = generic_pipe_buf_steal,
	64	.get = generic_pipe_buf_get
	65	};
	66	#else
f3bc08c5 MD	67	static const struct pipe_buf_operations ring_buffer_pipe_buf_ops = {
	68	.can_merge = 0,
	69	.map = generic_pipe_buf_map,
	70	.unmap = generic_pipe_buf_unmap,
	71	.confirm = generic_pipe_buf_confirm,
	72	.release = lib_ring_buffer_pipe_buf_release,
	73	.steal = generic_pipe_buf_steal,
4352d508	74	.get = generic_pipe_buf_get
f3bc08c5	75	};
4352d508	76	#endif
f3bc08c5 MD	77
	78	/*
	79	* Page release operation after splice pipe_to_file ends.
	80	*/
	81	static void lib_ring_buffer_page_release(struct splice_pipe_desc *spd,
	82	unsigned int i)
	83	{
	84	__free_page(spd->pages[i]);
	85	}
	86
	87	/*
	88	* subbuf_splice_actor - splice up to one subbuf's worth of data
	89	*/
	90	static int subbuf_splice_actor(struct file *in,
	91	loff_t *ppos,
	92	struct pipe_inode_info *pipe,
	93	size_t len,
d83004aa JD	94	unsigned int flags,
d83004aa JD	95	struct lib_ring_buffer *buf)
f3bc08c5	96	{
f3bc08c5	97	struct channel *chan = buf->backend.chan;
5a8fd222	98	const struct lib_ring_buffer_config *config = &chan->backend.config;
f3bc08c5 MD	99	unsigned int poff, subbuf_pages, nr_pages;
	100	struct page *pages[PIPE_DEF_BUFFERS];
	101	struct partial_page partial[PIPE_DEF_BUFFERS];
	102	struct splice_pipe_desc spd = {
	103	.pages = pages,
	104	.nr_pages = 0,
	105	.partial = partial,
e2833bed	106	#if (LINUX_VERSION_CODE < KERNEL_VERSION(4,12,0))
f3bc08c5	107	.flags = flags,
e2833bed	108	#endif
f3bc08c5 MD	109	.ops = &ring_buffer_pipe_buf_ops,
	110	.spd_release = lib_ring_buffer_page_release,
	111	};
88dfd899	112	unsigned long consumed_old, roffset;
f3bc08c5 MD	113	unsigned long bytes_avail;
	114
	115	/*
	116	* Check that a GET_SUBBUF ioctl has been done before.
	117	*/
	118	WARN_ON(atomic_long_read(&buf->active_readers) != 1);
	119	consumed_old = lib_ring_buffer_get_consumed(config, buf);
	120	consumed_old += *ppos;
f3bc08c5 MD	121
	122	/*
	123	* Adjust read len, if longer than what is available.
	124	* Max read size is 1 subbuffer due to get_subbuf/put_subbuf for
	125	* protection.
	126	*/
	127	bytes_avail = chan->backend.subbuf_size;
	128	WARN_ON(bytes_avail > chan->backend.buf_size);
	129	len = min_t(size_t, len, bytes_avail);
	130	subbuf_pages = bytes_avail >> PAGE_SHIFT;
	131	nr_pages = min_t(unsigned int, subbuf_pages, PIPE_DEF_BUFFERS);
	132	roffset = consumed_old & PAGE_MASK;
	133	poff = consumed_old & ~PAGE_MASK;
	134	printk_dbg(KERN_DEBUG "SPLICE actor len %zu pos %zd write_pos %ld\n",
	135	len, (ssize_t)*ppos, lib_ring_buffer_get_offset(config, buf));
	136
	137	for (; spd.nr_pages < nr_pages; spd.nr_pages++) {
	138	unsigned int this_len;
0112cb7b MD	139	unsigned long *pfnp, new_pfn;
0112cb7b MD	140	struct page *new_page;
f3bc08c5 MD	141	void **virt;
	142
	143	if (!len)
	144	break;
	145	printk_dbg(KERN_DEBUG "SPLICE actor loop len %zu roffset %ld\n",
	146	len, roffset);
	147
	148	/*
	149	* We have to replace the page we are moving into the splice
	150	* pipe.
	151	*/
	152	new_page = alloc_pages_node(cpu_to_node(max(buf->backend.cpu,
	153	0)),
	154	GFP_KERNEL \| __GFP_ZERO, 0);
	155	if (!new_page)
	156	break;
0112cb7b	157	new_pfn = page_to_pfn(new_page);
f3bc08c5	158	this_len = PAGE_SIZE - poff;
0112cb7b MD	159	pfnp = lib_ring_buffer_read_get_pfn(&buf->backend, roffset, &virt);
	160	spd.pages[spd.nr_pages] = pfn_to_page(*pfnp);
	161	*pfnp = new_pfn;
f3bc08c5 MD	162	*virt = page_address(new_page);
	163	spd.partial[spd.nr_pages].offset = poff;
	164	spd.partial[spd.nr_pages].len = this_len;
	165
	166	poff = 0;
	167	roffset += PAGE_SIZE;
	168	len -= this_len;
	169	}
	170
	171	if (!spd.nr_pages)
	172	return 0;
	173
90225db5	174	return wrapper_splice_to_pipe(pipe, &spd);
f3bc08c5 MD	175	}
	176
	177	ssize_t lib_ring_buffer_splice_read(struct file in, loff_t ppos,
	178	struct pipe_inode_info *pipe, size_t len,
d83004aa JD	179	unsigned int flags,
d83004aa JD	180	struct lib_ring_buffer *buf)
f3bc08c5	181	{
f3bc08c5	182	struct channel *chan = buf->backend.chan;
5a8fd222	183	const struct lib_ring_buffer_config *config = &chan->backend.config;
f3bc08c5 MD	184	ssize_t spliced;
	185	int ret;
	186
	187	if (config->output != RING_BUFFER_SPLICE)
	188	return -EINVAL;
	189
	190	/*
	191	* We require ppos and length to be page-aligned for performance reasons
	192	* (no page copy). Size is known using the ioctl
	193	* RING_BUFFER_GET_PADDED_SUBBUF_SIZE, which is page-size padded.
	194	* We fail when the ppos or len passed is not page-sized, because splice
	195	* is not allowed to copy more than the length passed as parameter (so
	196	* the ABI does not let us silently copy more than requested to include
	197	* padding).
	198	*/
	199	if (ppos != PAGE_ALIGN(ppos) \|\| len != PAGE_ALIGN(len))
	200	return -EINVAL;
	201
	202	ret = 0;
	203	spliced = 0;
	204
	205	printk_dbg(KERN_DEBUG "SPLICE read len %zu pos %zd\n", len,
	206	(ssize_t)*ppos);
	207	while (len && !spliced) {
d83004aa	208	ret = subbuf_splice_actor(in, ppos, pipe, len, flags, buf);
f3bc08c5 MD	209	printk_dbg(KERN_DEBUG "SPLICE read loop ret %d\n", ret);
	210	if (ret < 0)
	211	break;
	212	else if (!ret) {
	213	if (flags & SPLICE_F_NONBLOCK)
	214	ret = -EAGAIN;
	215	break;
	216	}
	217
	218	*ppos += ret;
	219	if (ret > len)
	220	len = 0;
	221	else
	222	len -= ret;
	223	spliced += ret;
	224	}
	225
	226	if (spliced)
	227	return spliced;
	228
	229	return ret;
	230	}
	231	EXPORT_SYMBOL_GPL(lib_ring_buffer_splice_read);
d83004aa JD	232
	233	ssize_t vfs_lib_ring_buffer_splice_read(struct file in, loff_t ppos,
	234	struct pipe_inode_info *pipe, size_t len,
	235	unsigned int flags)
	236	{
	237	struct lib_ring_buffer *buf = in->private_data;
	238
	239	return lib_ring_buffer_splice_read(in, ppos, pipe, len, flags, buf);
	240	}
	241	EXPORT_SYMBOL_GPL(vfs_lib_ring_buffer_splice_read);