[linux.git] / io_uring / tctx.c

// SPDX-License-Identifier: GPL-2.0
#include <linux/kernel.h>
#include <linux/errno.h>
#include <linux/file.h>
#include <linux/mm.h>
#include <linux/slab.h>
#include <linux/nospec.h>
#include <linux/io_uring.h>

#include <uapi/linux/io_uring.h>

#include "io_uring.h"
#include "tctx.h"

static struct io_wq *io_init_wq_offload(struct io_ring_ctx *ctx,
					struct task_struct *task)
{
	struct io_wq_hash *hash;
	struct io_wq_data data;
	unsigned int concurrency;

	mutex_lock(&ctx->uring_lock);
	hash = ctx->hash_map;
	if (!hash) {
		hash = kzalloc(sizeof(*hash), GFP_KERNEL);
		if (!hash) {
			mutex_unlock(&ctx->uring_lock);
			return ERR_PTR(-ENOMEM);
		}
		refcount_set(&hash->refs, 1);
		init_waitqueue_head(&hash->wait);
		ctx->hash_map = hash;
	}
	mutex_unlock(&ctx->uring_lock);

	data.hash = hash;
	data.task = task;
	data.free_work = io_wq_free_work;
	data.do_work = io_wq_submit_work;

	/* Do QD, or 4 * CPUS, whatever is smallest */
	concurrency = min(ctx->sq_entries, 4 * num_online_cpus());

	return io_wq_create(concurrency, &data);
}

void __io_uring_free(struct task_struct *tsk)
{
	struct io_uring_task *tctx = tsk->io_uring;

	WARN_ON_ONCE(!xa_empty(&tctx->xa));
	WARN_ON_ONCE(tctx->io_wq);
	WARN_ON_ONCE(tctx->cached_refs);

	percpu_counter_destroy(&tctx->inflight);
	kfree(tctx);
	tsk->io_uring = NULL;
}

__cold int io_uring_alloc_task_context(struct task_struct *task,
				       struct io_ring_ctx *ctx)
{
	struct io_uring_task *tctx;
	int ret;

	tctx = kzalloc(sizeof(*tctx), GFP_KERNEL);
	if (unlikely(!tctx))
		return -ENOMEM;

	ret = percpu_counter_init(&tctx->inflight, 0, GFP_KERNEL);
	if (unlikely(ret)) {
		kfree(tctx);
		return ret;
	}

	tctx->io_wq = io_init_wq_offload(ctx, task);
	if (IS_ERR(tctx->io_wq)) {
		ret = PTR_ERR(tctx->io_wq);
		percpu_counter_destroy(&tctx->inflight);
		kfree(tctx);
		return ret;
	}

	xa_init(&tctx->xa);
	init_waitqueue_head(&tctx->wait);
	atomic_set(&tctx->in_cancel, 0);
	atomic_set(&tctx->inflight_tracked, 0);
	task->io_uring = tctx;
	init_llist_head(&tctx->task_list);
	init_task_work(&tctx->task_work, tctx_task_work);
	return 0;
}

int __io_uring_add_tctx_node(struct io_ring_ctx *ctx)
{
	struct io_uring_task *tctx = current->io_uring;
	struct io_tctx_node *node;
	int ret;

	if (unlikely(!tctx)) {
		ret = io_uring_alloc_task_context(current, ctx);
		if (unlikely(ret))
			return ret;

		tctx = current->io_uring;
		if (ctx->iowq_limits_set) {
			unsigned int limits[2] = { ctx->iowq_limits[0],
						   ctx->iowq_limits[1], };

			ret = io_wq_max_workers(tctx->io_wq, limits);
			if (ret)
				return ret;
		}
	}
	if (!xa_load(&tctx->xa, (unsigned long)ctx)) {
		node = kmalloc(sizeof(*node), GFP_KERNEL);
		if (!node)
			return -ENOMEM;
		node->ctx = ctx;
		node->task = current;

		ret = xa_err(xa_store(&tctx->xa, (unsigned long)ctx,
					node, GFP_KERNEL));
		if (ret) {
			kfree(node);
			return ret;
		}

		mutex_lock(&ctx->uring_lock);
		list_add(&node->ctx_node, &ctx->tctx_list);
		mutex_unlock(&ctx->uring_lock);
	}
	return 0;
}

int __io_uring_add_tctx_node_from_submit(struct io_ring_ctx *ctx)
{
	int ret;

	if (ctx->flags & IORING_SETUP_SINGLE_ISSUER
	    && ctx->submitter_task != current)
		return -EEXIST;

	ret = __io_uring_add_tctx_node(ctx);
	if (ret)
		return ret;

	current->io_uring->last = ctx;
	return 0;
}

/*
 * Remove this io_uring_file -> task mapping.
 */
__cold void io_uring_del_tctx_node(unsigned long index)
{
	struct io_uring_task *tctx = current->io_uring;
	struct io_tctx_node *node;

	if (!tctx)
		return;
	node = xa_erase(&tctx->xa, index);
	if (!node)
		return;

	WARN_ON_ONCE(current != node->task);
	WARN_ON_ONCE(list_empty(&node->ctx_node));

	mutex_lock(&node->ctx->uring_lock);
	list_del(&node->ctx_node);
	mutex_unlock(&node->ctx->uring_lock);

	if (tctx->last == node->ctx)
		tctx->last = NULL;
	kfree(node);
}

__cold void io_uring_clean_tctx(struct io_uring_task *tctx)
{
	struct io_wq *wq = tctx->io_wq;
	struct io_tctx_node *node;
	unsigned long index;

	xa_for_each(&tctx->xa, index, node) {
		io_uring_del_tctx_node(index);
		cond_resched();
	}
	if (wq) {
		/*
		 * Must be after io_uring_del_tctx_node() (removes nodes under
		 * uring_lock) to avoid race with io_uring_try_cancel_iowq().
		 */
		io_wq_put_and_exit(wq);
		tctx->io_wq = NULL;
	}
}

void io_uring_unreg_ringfd(void)
{
	struct io_uring_task *tctx = current->io_uring;
	int i;

	for (i = 0; i < IO_RINGFD_REG_MAX; i++) {
		if (tctx->registered_rings[i]) {
			fput(tctx->registered_rings[i]);
			tctx->registered_rings[i] = NULL;
		}
	}
}

static int io_ring_add_registered_fd(struct io_uring_task *tctx, int fd,
				     int start, int end)
{
	struct file *file;
	int offset;

	for (offset = start; offset < end; offset++) {
		offset = array_index_nospec(offset, IO_RINGFD_REG_MAX);
		if (tctx->registered_rings[offset])
			continue;

		file = fget(fd);
		if (!file) {
			return -EBADF;
		} else if (!io_is_uring_fops(file)) {
			fput(file);
			return -EOPNOTSUPP;
		}
		tctx->registered_rings[offset] = file;
		return offset;
	}

	return -EBUSY;
}

/*
 * Register a ring fd to avoid fdget/fdput for each io_uring_enter()
 * invocation. User passes in an array of struct io_uring_rsrc_update
 * with ->data set to the ring_fd, and ->offset given for the desired
 * index. If no index is desired, application may set ->offset == -1U
 * and we'll find an available index. Returns number of entries
 * successfully processed, or < 0 on error if none were processed.
 */
int io_ringfd_register(struct io_ring_ctx *ctx, void __user *__arg,
		       unsigned nr_args)
{
	struct io_uring_rsrc_update __user *arg = __arg;
	struct io_uring_rsrc_update reg;
	struct io_uring_task *tctx;
	int ret, i;

	if (!nr_args || nr_args > IO_RINGFD_REG_MAX)
		return -EINVAL;

	mutex_unlock(&ctx->uring_lock);
	ret = __io_uring_add_tctx_node(ctx);
	mutex_lock(&ctx->uring_lock);
	if (ret)
		return ret;

	tctx = current->io_uring;
	for (i = 0; i < nr_args; i++) {
		int start, end;

		if (copy_from_user(&reg, &arg[i], sizeof(reg))) {
			ret = -EFAULT;
			break;
		}

		if (reg.resv) {
			ret = -EINVAL;
			break;
		}

		if (reg.offset == -1U) {
			start = 0;
			end = IO_RINGFD_REG_MAX;
		} else {
			if (reg.offset >= IO_RINGFD_REG_MAX) {
				ret = -EINVAL;
				break;
			}
			start = reg.offset;
			end = start + 1;
		}

		ret = io_ring_add_registered_fd(tctx, reg.data, start, end);
		if (ret < 0)
			break;

		reg.offset = ret;
		if (copy_to_user(&arg[i], &reg, sizeof(reg))) {
			fput(tctx->registered_rings[reg.offset]);
			tctx->registered_rings[reg.offset] = NULL;
			ret = -EFAULT;
			break;
		}
	}

	return i ? i : ret;
}

int io_ringfd_unregister(struct io_ring_ctx *ctx, void __user *__arg,
			 unsigned nr_args)
{
	struct io_uring_rsrc_update __user *arg = __arg;
	struct io_uring_task *tctx = current->io_uring;
	struct io_uring_rsrc_update reg;
	int ret = 0, i;

	if (!nr_args || nr_args > IO_RINGFD_REG_MAX)
		return -EINVAL;
	if (!tctx)
		return 0;

	for (i = 0; i < nr_args; i++) {
		if (copy_from_user(&reg, &arg[i], sizeof(reg))) {
			ret = -EFAULT;
			break;
		}
		if (reg.resv || reg.data || reg.offset >= IO_RINGFD_REG_MAX) {
			ret = -EINVAL;
			break;
		}

		reg.offset = array_index_nospec(reg.offset, IO_RINGFD_REG_MAX);
		if (tctx->registered_rings[reg.offset]) {
			fput(tctx->registered_rings[reg.offset]);
			tctx->registered_rings[reg.offset] = NULL;
		}
	}

	return i ? i : ret;
}
Commit	Line	Data
c9f06aa7 JA	1	// SPDX-License-Identifier: GPL-2.0
	2	#include <linux/kernel.h>
	3	#include <linux/errno.h>
	4	#include <linux/file.h>
	5	#include <linux/mm.h>
	6	#include <linux/slab.h>
	7	#include <linux/nospec.h>
	8	#include <linux/io_uring.h>
	9
	10	#include <uapi/linux/io_uring.h>
	11
c9f06aa7 JA	12	#include "io_uring.h"
	13	#include "tctx.h"
	14
	15	static struct io_wq io_init_wq_offload(struct io_ring_ctx ctx,
	16	struct task_struct *task)
	17	{
	18	struct io_wq_hash *hash;
	19	struct io_wq_data data;
	20	unsigned int concurrency;
	21
	22	mutex_lock(&ctx->uring_lock);
	23	hash = ctx->hash_map;
	24	if (!hash) {
	25	hash = kzalloc(sizeof(*hash), GFP_KERNEL);
	26	if (!hash) {
	27	mutex_unlock(&ctx->uring_lock);
	28	return ERR_PTR(-ENOMEM);
	29	}
	30	refcount_set(&hash->refs, 1);
	31	init_waitqueue_head(&hash->wait);
	32	ctx->hash_map = hash;
	33	}
	34	mutex_unlock(&ctx->uring_lock);
	35
	36	data.hash = hash;
	37	data.task = task;
	38	data.free_work = io_wq_free_work;
	39	data.do_work = io_wq_submit_work;
	40
	41	/* Do QD, or 4 * CPUS, whatever is smallest */
	42	concurrency = min(ctx->sq_entries, 4 * num_online_cpus());
	43
	44	return io_wq_create(concurrency, &data);
	45	}
	46
	47	void __io_uring_free(struct task_struct *tsk)
	48	{
	49	struct io_uring_task *tctx = tsk->io_uring;
	50
	51	WARN_ON_ONCE(!xa_empty(&tctx->xa));
	52	WARN_ON_ONCE(tctx->io_wq);
	53	WARN_ON_ONCE(tctx->cached_refs);
	54
c9f06aa7 JA	55	percpu_counter_destroy(&tctx->inflight);
	56	kfree(tctx);
	57	tsk->io_uring = NULL;
	58	}
	59
	60	__cold int io_uring_alloc_task_context(struct task_struct *task,
	61	struct io_ring_ctx *ctx)
	62	{
	63	struct io_uring_task *tctx;
	64	int ret;
	65
	66	tctx = kzalloc(sizeof(*tctx), GFP_KERNEL);
	67	if (unlikely(!tctx))
	68	return -ENOMEM;
	69
c9f06aa7 JA	70	ret = percpu_counter_init(&tctx->inflight, 0, GFP_KERNEL);
c9f06aa7 JA	71	if (unlikely(ret)) {
c9f06aa7 JA	72	kfree(tctx);
	73	return ret;
	74	}
	75
	76	tctx->io_wq = io_init_wq_offload(ctx, task);
	77	if (IS_ERR(tctx->io_wq)) {
	78	ret = PTR_ERR(tctx->io_wq);
	79	percpu_counter_destroy(&tctx->inflight);
c9f06aa7 JA	80	kfree(tctx);
	81	return ret;
	82	}
	83
	84	xa_init(&tctx->xa);
	85	init_waitqueue_head(&tctx->wait);
8d664282	86	atomic_set(&tctx->in_cancel, 0);
c9f06aa7 JA	87	atomic_set(&tctx->inflight_tracked, 0);
c9f06aa7 JA	88	task->io_uring = tctx;
f88262e6	89	init_llist_head(&tctx->task_list);
c9f06aa7 JA	90	init_task_work(&tctx->task_work, tctx_task_work);
	91	return 0;
	92	}
	93
97c96e9f	94	int __io_uring_add_tctx_node(struct io_ring_ctx *ctx)
c9f06aa7 JA	95	{
	96	struct io_uring_task *tctx = current->io_uring;
	97	struct io_tctx_node *node;
	98	int ret;
	99
	100	if (unlikely(!tctx)) {
	101	ret = io_uring_alloc_task_context(current, ctx);
	102	if (unlikely(ret))
	103	return ret;
	104
	105	tctx = current->io_uring;
	106	if (ctx->iowq_limits_set) {
	107	unsigned int limits[2] = { ctx->iowq_limits[0],
	108	ctx->iowq_limits[1], };
	109
	110	ret = io_wq_max_workers(tctx->io_wq, limits);
	111	if (ret)
	112	return ret;
	113	}
	114	}
	115	if (!xa_load(&tctx->xa, (unsigned long)ctx)) {
	116	node = kmalloc(sizeof(*node), GFP_KERNEL);
	117	if (!node)
	118	return -ENOMEM;
	119	node->ctx = ctx;
	120	node->task = current;
	121
	122	ret = xa_err(xa_store(&tctx->xa, (unsigned long)ctx,
	123	node, GFP_KERNEL));
	124	if (ret) {
	125	kfree(node);
	126	return ret;
	127	}
	128
	129	mutex_lock(&ctx->uring_lock);
	130	list_add(&node->ctx_node, &ctx->tctx_list);
	131	mutex_unlock(&ctx->uring_lock);
	132	}
97c96e9f DY	133	return 0;
	134	}
	135
	136	int __io_uring_add_tctx_node_from_submit(struct io_ring_ctx *ctx)
	137	{
	138	int ret;
	139
4add705e DY	140	if (ctx->flags & IORING_SETUP_SINGLE_ISSUER
	141	&& ctx->submitter_task != current)
	142	return -EEXIST;
97c96e9f DY	143
	144	ret = __io_uring_add_tctx_node(ctx);
	145	if (ret)
	146	return ret;
	147
	148	current->io_uring->last = ctx;
c9f06aa7 JA	149	return 0;
	150	}
	151
	152	/*
	153	* Remove this io_uring_file -> task mapping.
	154	*/
	155	__cold void io_uring_del_tctx_node(unsigned long index)
	156	{
	157	struct io_uring_task *tctx = current->io_uring;
	158	struct io_tctx_node *node;
	159
	160	if (!tctx)
	161	return;
	162	node = xa_erase(&tctx->xa, index);
	163	if (!node)
	164	return;
	165
	166	WARN_ON_ONCE(current != node->task);
	167	WARN_ON_ONCE(list_empty(&node->ctx_node));
	168
	169	mutex_lock(&node->ctx->uring_lock);
	170	list_del(&node->ctx_node);
	171	mutex_unlock(&node->ctx->uring_lock);
	172
	173	if (tctx->last == node->ctx)
	174	tctx->last = NULL;
	175	kfree(node);
	176	}
	177
	178	__cold void io_uring_clean_tctx(struct io_uring_task *tctx)
	179	{
	180	struct io_wq *wq = tctx->io_wq;
	181	struct io_tctx_node *node;
	182	unsigned long index;
	183
	184	xa_for_each(&tctx->xa, index, node) {
	185	io_uring_del_tctx_node(index);
	186	cond_resched();
	187	}
	188	if (wq) {
	189	/*
	190	* Must be after io_uring_del_tctx_node() (removes nodes under
	191	* uring_lock) to avoid race with io_uring_try_cancel_iowq().
	192	*/
	193	io_wq_put_and_exit(wq);
	194	tctx->io_wq = NULL;
	195	}
	196	}
	197
	198	void io_uring_unreg_ringfd(void)
	199	{
	200	struct io_uring_task *tctx = current->io_uring;
	201	int i;
	202
	203	for (i = 0; i < IO_RINGFD_REG_MAX; i++) {
	204	if (tctx->registered_rings[i]) {
	205	fput(tctx->registered_rings[i]);
	206	tctx->registered_rings[i] = NULL;
	207	}
	208	}
	209	}
	210
	211	static int io_ring_add_registered_fd(struct io_uring_task *tctx, int fd,
	212	int start, int end)
213	{
214	struct file *file;
215	int offset;
216
217	for (offset = start; offset < end; offset++) {
218	offset = array_index_nospec(offset, IO_RINGFD_REG_MAX);
219	if (tctx->registered_rings[offset])
220	continue;
221
222	file = fget(fd);
223	if (!file) {
224	return -EBADF;
225	} else if (!io_is_uring_fops(file)) {
226	fput(file);
227	return -EOPNOTSUPP;
228	}
229	tctx->registered_rings[offset] = file;
230	return offset;
231	}
232
233	return -EBUSY;
234	}
235
236	/*
237	* Register a ring fd to avoid fdget/fdput for each io_uring_enter()
238	* invocation. User passes in an array of struct io_uring_rsrc_update
239	* with ->data set to the ring_fd, and ->offset given for the desired
240	* index. If no index is desired, application may set ->offset == -1U
241	* and we'll find an available index. Returns number of entries
242	* successfully processed, or < 0 on error if none were processed.
243	*/
244	int io_ringfd_register(struct io_ring_ctx ctx, void __user __arg,
245	unsigned nr_args)
246	{
247	struct io_uring_rsrc_update __user *arg = __arg;
248	struct io_uring_rsrc_update reg;
249	struct io_uring_task *tctx;
250	int ret, i;
251
252	if (!nr_args \|\| nr_args > IO_RINGFD_REG_MAX)
253	return -EINVAL;
254
255	mutex_unlock(&ctx->uring_lock);
97c96e9f	256	ret = __io_uring_add_tctx_node(ctx);
c9f06aa7 JA	257	mutex_lock(&ctx->uring_lock);
	258	if (ret)
	259	return ret;
	260
	261	tctx = current->io_uring;
	262	for (i = 0; i < nr_args; i++) {
	263	int start, end;
	264
	265	if (copy_from_user(&reg, &arg[i], sizeof(reg))) {
	266	ret = -EFAULT;
	267	break;
	268	}
	269
	270	if (reg.resv) {
	271	ret = -EINVAL;
	272	break;
	273	}
	274
	275	if (reg.offset == -1U) {
	276	start = 0;
	277	end = IO_RINGFD_REG_MAX;
	278	} else {
	279	if (reg.offset >= IO_RINGFD_REG_MAX) {
	280	ret = -EINVAL;
	281	break;
	282	}
	283	start = reg.offset;
	284	end = start + 1;
	285	}
	286
	287	ret = io_ring_add_registered_fd(tctx, reg.data, start, end);
	288	if (ret < 0)
	289	break;
	290
	291	reg.offset = ret;
	292	if (copy_to_user(&arg[i], &reg, sizeof(reg))) {
	293	fput(tctx->registered_rings[reg.offset]);
	294	tctx->registered_rings[reg.offset] = NULL;
	295	ret = -EFAULT;
	296	break;
	297	}
	298	}
	299
	300	return i ? i : ret;
	301	}
	302
	303	int io_ringfd_unregister(struct io_ring_ctx ctx, void __user __arg,
	304	unsigned nr_args)
	305	{
	306	struct io_uring_rsrc_update __user *arg = __arg;
	307	struct io_uring_task *tctx = current->io_uring;
	308	struct io_uring_rsrc_update reg;
	309	int ret = 0, i;
	310
	311	if (!nr_args \|\| nr_args > IO_RINGFD_REG_MAX)
	312	return -EINVAL;
	313	if (!tctx)
	314	return 0;
	315
	316	for (i = 0; i < nr_args; i++) {
	317	if (copy_from_user(&reg, &arg[i], sizeof(reg))) {
	318	ret = -EFAULT;
	319	break;
	320	}
321	if (reg.resv \|\| reg.data \|\| reg.offset >= IO_RINGFD_REG_MAX) {
322	ret = -EINVAL;
323	break;
324	}
325
326	reg.offset = array_index_nospec(reg.offset, IO_RINGFD_REG_MAX);
327	if (tctx->registered_rings[reg.offset]) {
328	fput(tctx->registered_rings[reg.offset]);
329	tctx->registered_rings[reg.offset] = NULL;
330	}
331	}
332
333	return i ? i : ret;
334	}