[linux-block.git] / fs / cachefiles / io.c

// SPDX-License-Identifier: GPL-2.0-or-later
/* kiocb-using read/write
 *
 * Copyright (C) 2021 Red Hat, Inc. All Rights Reserved.
 * Written by David Howells (dhowells@redhat.com)
 */

#include <linux/mount.h>
#include <linux/slab.h>
#include <linux/file.h>
#include <linux/uio.h>
#include <linux/falloc.h>
#include <linux/sched/mm.h>
#include <trace/events/fscache.h>
#include "internal.h"

struct cachefiles_kiocb {
	struct kiocb		iocb;
	refcount_t		ki_refcnt;
	loff_t			start;
	union {
		size_t		skipped;
		size_t		len;
	};
	struct cachefiles_object *object;
	netfs_io_terminated_t	term_func;
	void			*term_func_priv;
	bool			was_async;
	unsigned int		inval_counter;	/* Copy of cookie->inval_counter */
	u64			b_writing;
};

static inline void cachefiles_put_kiocb(struct cachefiles_kiocb *ki)
{
	if (refcount_dec_and_test(&ki->ki_refcnt)) {
		cachefiles_put_object(ki->object, cachefiles_obj_put_ioreq);
		fput(ki->iocb.ki_filp);
		kfree(ki);
	}
}

/*
 * Handle completion of a read from the cache.
 */
static void cachefiles_read_complete(struct kiocb *iocb, long ret)
{
	struct cachefiles_kiocb *ki = container_of(iocb, struct cachefiles_kiocb, iocb);
	struct inode *inode = file_inode(ki->iocb.ki_filp);

	_enter("%ld", ret);

	if (ret < 0)
		trace_cachefiles_io_error(ki->object, inode, ret,
					  cachefiles_trace_read_error);

	if (ki->term_func) {
		if (ret >= 0) {
			if (ki->object->cookie->inval_counter == ki->inval_counter)
				ki->skipped += ret;
			else
				ret = -ESTALE;
		}

		ki->term_func(ki->term_func_priv, ret, ki->was_async);
	}

	cachefiles_put_kiocb(ki);
}

/*
 * Initiate a read from the cache.
 */
static int cachefiles_read(struct netfs_cache_resources *cres,
			   loff_t start_pos,
			   struct iov_iter *iter,
			   enum netfs_read_from_hole read_hole,
			   netfs_io_terminated_t term_func,
			   void *term_func_priv)
{
	struct cachefiles_object *object;
	struct cachefiles_kiocb *ki;
	struct file *file;
	unsigned int old_nofs;
	ssize_t ret = -ENOBUFS;
	size_t len = iov_iter_count(iter), skipped = 0;

	if (!fscache_wait_for_operation(cres, FSCACHE_WANT_READ))
		goto presubmission_error;

	fscache_count_read();
	object = cachefiles_cres_object(cres);
	file = cachefiles_cres_file(cres);

	_enter("%pD,%li,%llx,%zx/%llx",
	       file, file_inode(file)->i_ino, start_pos, len,
	       i_size_read(file_inode(file)));

	/* If the caller asked us to seek for data before doing the read, then
	 * we should do that now.  If we find a gap, we fill it with zeros.
	 */
	if (read_hole != NETFS_READ_HOLE_IGNORE) {
		loff_t off = start_pos, off2;

		off2 = cachefiles_inject_read_error();
		if (off2 == 0)
			off2 = vfs_llseek(file, off, SEEK_DATA);
		if (off2 < 0 && off2 >= (loff_t)-MAX_ERRNO && off2 != -ENXIO) {
			skipped = 0;
			ret = off2;
			goto presubmission_error;
		}

		if (off2 == -ENXIO || off2 >= start_pos + len) {
			/* The region is beyond the EOF or there's no more data
			 * in the region, so clear the rest of the buffer and
			 * return success.
			 */
			ret = -ENODATA;
			if (read_hole == NETFS_READ_HOLE_FAIL)
				goto presubmission_error;

			iov_iter_zero(len, iter);
			skipped = len;
			ret = 0;
			goto presubmission_error;
		}

		skipped = off2 - off;
		iov_iter_zero(skipped, iter);
	}

	ret = -ENOMEM;
	ki = kzalloc(sizeof(struct cachefiles_kiocb), GFP_KERNEL);
	if (!ki)
		goto presubmission_error;

	refcount_set(&ki->ki_refcnt, 2);
	ki->iocb.ki_filp	= file;
	ki->iocb.ki_pos		= start_pos + skipped;
	ki->iocb.ki_flags	= IOCB_DIRECT;
	ki->iocb.ki_hint	= ki_hint_validate(file_write_hint(file));
	ki->iocb.ki_ioprio	= get_current_ioprio();
	ki->skipped		= skipped;
	ki->object		= object;
	ki->inval_counter	= cres->inval_counter;
	ki->term_func		= term_func;
	ki->term_func_priv	= term_func_priv;
	ki->was_async		= true;

	if (ki->term_func)
		ki->iocb.ki_complete = cachefiles_read_complete;

	get_file(ki->iocb.ki_filp);
	cachefiles_grab_object(object, cachefiles_obj_get_ioreq);

	trace_cachefiles_read(object, file_inode(file), ki->iocb.ki_pos, len - skipped);
	old_nofs = memalloc_nofs_save();
	ret = cachefiles_inject_read_error();
	if (ret == 0)
		ret = vfs_iocb_iter_read(file, &ki->iocb, iter);
	memalloc_nofs_restore(old_nofs);
	switch (ret) {
	case -EIOCBQUEUED:
		goto in_progress;

	case -ERESTARTSYS:
	case -ERESTARTNOINTR:
	case -ERESTARTNOHAND:
	case -ERESTART_RESTARTBLOCK:
		/* There's no easy way to restart the syscall since other AIO's
		 * may be already running. Just fail this IO with EINTR.
		 */
		ret = -EINTR;
		fallthrough;
	default:
		ki->was_async = false;
		cachefiles_read_complete(&ki->iocb, ret);
		if (ret > 0)
			ret = 0;
		break;
	}

in_progress:
	cachefiles_put_kiocb(ki);
	_leave(" = %zd", ret);
	return ret;

presubmission_error:
	if (term_func)
		term_func(term_func_priv, ret < 0 ? ret : skipped, false);
	return ret;
}

/*
 * Query the occupancy of the cache in a region, returning where the next chunk
 * of data starts and how long it is.
 */
static int cachefiles_query_occupancy(struct netfs_cache_resources *cres,
				      loff_t start, size_t len, size_t granularity,
				      loff_t *_data_start, size_t *_data_len)
{
	struct cachefiles_object *object;
	struct file *file;
	loff_t off, off2;

	*_data_start = -1;
	*_data_len = 0;

	if (!fscache_wait_for_operation(cres, FSCACHE_WANT_READ))
		return -ENOBUFS;

	object = cachefiles_cres_object(cres);
	file = cachefiles_cres_file(cres);
	granularity = max_t(size_t, object->volume->cache->bsize, granularity);

	_enter("%pD,%li,%llx,%zx/%llx",
	       file, file_inode(file)->i_ino, start, len,
	       i_size_read(file_inode(file)));

	off = cachefiles_inject_read_error();
	if (off == 0)
		off = vfs_llseek(file, start, SEEK_DATA);
	if (off == -ENXIO)
		return -ENODATA; /* Beyond EOF */
	if (off < 0 && off >= (loff_t)-MAX_ERRNO)
		return -ENOBUFS; /* Error. */
	if (round_up(off, granularity) >= start + len)
		return -ENODATA; /* No data in range */

	off2 = cachefiles_inject_read_error();
	if (off2 == 0)
		off2 = vfs_llseek(file, off, SEEK_HOLE);
	if (off2 == -ENXIO)
		return -ENODATA; /* Beyond EOF */
	if (off2 < 0 && off2 >= (loff_t)-MAX_ERRNO)
		return -ENOBUFS; /* Error. */

	/* Round away partial blocks */
	off = round_up(off, granularity);
	off2 = round_down(off2, granularity);
	if (off2 <= off)
		return -ENODATA;

	*_data_start = off;
	if (off2 > start + len)
		*_data_len = len;
	else
		*_data_len = off2 - off;
	return 0;
}

/*
 * Handle completion of a write to the cache.
 */
static void cachefiles_write_complete(struct kiocb *iocb, long ret)
{
	struct cachefiles_kiocb *ki = container_of(iocb, struct cachefiles_kiocb, iocb);
	struct cachefiles_object *object = ki->object;
	struct inode *inode = file_inode(ki->iocb.ki_filp);

	_enter("%ld", ret);

	/* Tell lockdep we inherited freeze protection from submission thread */
	__sb_writers_acquired(inode->i_sb, SB_FREEZE_WRITE);
	__sb_end_write(inode->i_sb, SB_FREEZE_WRITE);

	if (ret < 0)
		trace_cachefiles_io_error(object, inode, ret,
					  cachefiles_trace_write_error);

	atomic_long_sub(ki->b_writing, &object->volume->cache->b_writing);
	set_bit(FSCACHE_COOKIE_HAVE_DATA, &object->cookie->flags);
	if (ki->term_func)
		ki->term_func(ki->term_func_priv, ret, ki->was_async);
	cachefiles_put_kiocb(ki);
}

/*
 * Initiate a write to the cache.
 */
static int cachefiles_write(struct netfs_cache_resources *cres,
			    loff_t start_pos,
			    struct iov_iter *iter,
			    netfs_io_terminated_t term_func,
			    void *term_func_priv)
{
	struct cachefiles_object *object;
	struct cachefiles_cache *cache;
	struct cachefiles_kiocb *ki;
	struct inode *inode;
	struct file *file;
	unsigned int old_nofs;
	ssize_t ret = -ENOBUFS;
	size_t len = iov_iter_count(iter);

	if (!fscache_wait_for_operation(cres, FSCACHE_WANT_WRITE))
		goto presubmission_error;
	fscache_count_write();
	object = cachefiles_cres_object(cres);
	cache = object->volume->cache;
	file = cachefiles_cres_file(cres);

	_enter("%pD,%li,%llx,%zx/%llx",
	       file, file_inode(file)->i_ino, start_pos, len,
	       i_size_read(file_inode(file)));

	ret = -ENOMEM;
	ki = kzalloc(sizeof(struct cachefiles_kiocb), GFP_KERNEL);
	if (!ki)
		goto presubmission_error;

	refcount_set(&ki->ki_refcnt, 2);
	ki->iocb.ki_filp	= file;
	ki->iocb.ki_pos		= start_pos;
	ki->iocb.ki_flags	= IOCB_DIRECT | IOCB_WRITE;
	ki->iocb.ki_hint	= ki_hint_validate(file_write_hint(file));
	ki->iocb.ki_ioprio	= get_current_ioprio();
	ki->object		= object;
	ki->inval_counter	= cres->inval_counter;
	ki->start		= start_pos;
	ki->len			= len;
	ki->term_func		= term_func;
	ki->term_func_priv	= term_func_priv;
	ki->was_async		= true;
	ki->b_writing		= (len + (1 << cache->bshift) - 1) >> cache->bshift;

	if (ki->term_func)
		ki->iocb.ki_complete = cachefiles_write_complete;
	atomic_long_add(ki->b_writing, &cache->b_writing);

	/* Open-code file_start_write here to grab freeze protection, which
	 * will be released by another thread in aio_complete_rw().  Fool
	 * lockdep by telling it the lock got released so that it doesn't
	 * complain about the held lock when we return to userspace.
	 */
	inode = file_inode(file);
	__sb_start_write(inode->i_sb, SB_FREEZE_WRITE);
	__sb_writers_release(inode->i_sb, SB_FREEZE_WRITE);

	get_file(ki->iocb.ki_filp);
	cachefiles_grab_object(object, cachefiles_obj_get_ioreq);

	trace_cachefiles_write(object, inode, ki->iocb.ki_pos, len);
	old_nofs = memalloc_nofs_save();
	ret = cachefiles_inject_write_error();
	if (ret == 0)
		ret = vfs_iocb_iter_write(file, &ki->iocb, iter);
	memalloc_nofs_restore(old_nofs);
	switch (ret) {
	case -EIOCBQUEUED:
		goto in_progress;

	case -ERESTARTSYS:
	case -ERESTARTNOINTR:
	case -ERESTARTNOHAND:
	case -ERESTART_RESTARTBLOCK:
		/* There's no easy way to restart the syscall since other AIO's
		 * may be already running. Just fail this IO with EINTR.
		 */
		ret = -EINTR;
		fallthrough;
	default:
		ki->was_async = false;
		cachefiles_write_complete(&ki->iocb, ret);
		if (ret > 0)
			ret = 0;
		break;
	}

in_progress:
	cachefiles_put_kiocb(ki);
	_leave(" = %zd", ret);
	return ret;

presubmission_error:
	if (term_func)
		term_func(term_func_priv, ret, false);
	return ret;
}

/*
 * Prepare a read operation, shortening it to a cached/uncached
 * boundary as appropriate.
 */
static enum netfs_io_source cachefiles_prepare_read(struct netfs_io_subrequest *subreq,
						      loff_t i_size)
{
	enum cachefiles_prepare_read_trace why;
	struct netfs_io_request *rreq = subreq->rreq;
	struct netfs_cache_resources *cres = &rreq->cache_resources;
	struct cachefiles_object *object;
	struct cachefiles_cache *cache;
	struct fscache_cookie *cookie = fscache_cres_cookie(cres);
	const struct cred *saved_cred;
	struct file *file = cachefiles_cres_file(cres);
	enum netfs_io_source ret = NETFS_DOWNLOAD_FROM_SERVER;
	loff_t off, to;
	ino_t ino = file ? file_inode(file)->i_ino : 0;

	_enter("%zx @%llx/%llx", subreq->len, subreq->start, i_size);

	if (subreq->start >= i_size) {
		ret = NETFS_FILL_WITH_ZEROES;
		why = cachefiles_trace_read_after_eof;
		goto out_no_object;
	}

	if (test_bit(FSCACHE_COOKIE_NO_DATA_TO_READ, &cookie->flags)) {
		__set_bit(NETFS_SREQ_WRITE_TO_CACHE, &subreq->flags);
		why = cachefiles_trace_read_no_data;
		goto out_no_object;
	}

	/* The object and the file may be being created in the background. */
	if (!file) {
		why = cachefiles_trace_read_no_file;
		if (!fscache_wait_for_operation(cres, FSCACHE_WANT_READ))
			goto out_no_object;
		file = cachefiles_cres_file(cres);
		if (!file)
			goto out_no_object;
		ino = file_inode(file)->i_ino;
	}

	object = cachefiles_cres_object(cres);
	cache = object->volume->cache;
	cachefiles_begin_secure(cache, &saved_cred);

	off = cachefiles_inject_read_error();
	if (off == 0)
		off = vfs_llseek(file, subreq->start, SEEK_DATA);
	if (off < 0 && off >= (loff_t)-MAX_ERRNO) {
		if (off == (loff_t)-ENXIO) {
			why = cachefiles_trace_read_seek_nxio;
			goto download_and_store;
		}
		trace_cachefiles_io_error(object, file_inode(file), off,
					  cachefiles_trace_seek_error);
		why = cachefiles_trace_read_seek_error;
		goto out;
	}

	if (off >= subreq->start + subreq->len) {
		why = cachefiles_trace_read_found_hole;
		goto download_and_store;
	}

	if (off > subreq->start) {
		off = round_up(off, cache->bsize);
		subreq->len = off - subreq->start;
		why = cachefiles_trace_read_found_part;
		goto download_and_store;
	}

	to = cachefiles_inject_read_error();
	if (to == 0)
		to = vfs_llseek(file, subreq->start, SEEK_HOLE);
	if (to < 0 && to >= (loff_t)-MAX_ERRNO) {
		trace_cachefiles_io_error(object, file_inode(file), to,
					  cachefiles_trace_seek_error);
		why = cachefiles_trace_read_seek_error;
		goto out;
	}

	if (to < subreq->start + subreq->len) {
		if (subreq->start + subreq->len >= i_size)
			to = round_up(to, cache->bsize);
		else
			to = round_down(to, cache->bsize);
		subreq->len = to - subreq->start;
	}

	why = cachefiles_trace_read_have_data;
	ret = NETFS_READ_FROM_CACHE;
	goto out;

download_and_store:
	__set_bit(NETFS_SREQ_WRITE_TO_CACHE, &subreq->flags);
out:
	cachefiles_end_secure(cache, saved_cred);
out_no_object:
	trace_cachefiles_prep_read(subreq, ret, why, ino);
	return ret;
}

/*
 * Prepare for a write to occur.
 */
static int __cachefiles_prepare_write(struct netfs_cache_resources *cres,
				      loff_t *_start, size_t *_len, loff_t i_size,
				      bool no_space_allocated_yet)
{
	struct cachefiles_object *object = cachefiles_cres_object(cres);
	struct cachefiles_cache *cache = object->volume->cache;
	struct file *file = cachefiles_cres_file(cres);
	loff_t start = *_start, pos;
	size_t len = *_len, down;
	int ret;

	/* Round to DIO size */
	down = start - round_down(start, PAGE_SIZE);
	*_start = start - down;
	*_len = round_up(down + len, PAGE_SIZE);

	/* We need to work out whether there's sufficient disk space to perform
	 * the write - but we can skip that check if we have space already
	 * allocated.
	 */
	if (no_space_allocated_yet)
		goto check_space;

	pos = cachefiles_inject_read_error();
	if (pos == 0)
		pos = vfs_llseek(file, *_start, SEEK_DATA);
	if (pos < 0 && pos >= (loff_t)-MAX_ERRNO) {
		if (pos == -ENXIO)
			goto check_space; /* Unallocated tail */
		trace_cachefiles_io_error(object, file_inode(file), pos,
					  cachefiles_trace_seek_error);
		return pos;
	}
	if ((u64)pos >= (u64)*_start + *_len)
		goto check_space; /* Unallocated region */

	/* We have a block that's at least partially filled - if we're low on
	 * space, we need to see if it's fully allocated.  If it's not, we may
	 * want to cull it.
	 */
	if (cachefiles_has_space(cache, 0, *_len / PAGE_SIZE,
				 cachefiles_has_space_check) == 0)
		return 0; /* Enough space to simply overwrite the whole block */

	pos = cachefiles_inject_read_error();
	if (pos == 0)
		pos = vfs_llseek(file, *_start, SEEK_HOLE);
	if (pos < 0 && pos >= (loff_t)-MAX_ERRNO) {
		trace_cachefiles_io_error(object, file_inode(file), pos,
					  cachefiles_trace_seek_error);
		return pos;
	}
	if ((u64)pos >= (u64)*_start + *_len)
		return 0; /* Fully allocated */

	/* Partially allocated, but insufficient space: cull. */
	fscache_count_no_write_space();
	ret = cachefiles_inject_remove_error();
	if (ret == 0)
		ret = vfs_fallocate(file, FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
				    *_start, *_len);
	if (ret < 0) {
		trace_cachefiles_io_error(object, file_inode(file), ret,
					  cachefiles_trace_fallocate_error);
		cachefiles_io_error_obj(object,
					"CacheFiles: fallocate failed (%d)\n", ret);
		ret = -EIO;
	}

	return ret;

check_space:
	return cachefiles_has_space(cache, 0, *_len / PAGE_SIZE,
				    cachefiles_has_space_for_write);
}

static int cachefiles_prepare_write(struct netfs_cache_resources *cres,
				    loff_t *_start, size_t *_len, loff_t i_size,
				    bool no_space_allocated_yet)
{
	struct cachefiles_object *object = cachefiles_cres_object(cres);
	struct cachefiles_cache *cache = object->volume->cache;
	const struct cred *saved_cred;
	int ret;

	if (!cachefiles_cres_file(cres)) {
		if (!fscache_wait_for_operation(cres, FSCACHE_WANT_WRITE))
			return -ENOBUFS;
		if (!cachefiles_cres_file(cres))
			return -ENOBUFS;
	}

	cachefiles_begin_secure(cache, &saved_cred);
	ret = __cachefiles_prepare_write(cres, _start, _len, i_size,
					 no_space_allocated_yet);
	cachefiles_end_secure(cache, saved_cred);
	return ret;
}

/*
 * Clean up an operation.
 */
static void cachefiles_end_operation(struct netfs_cache_resources *cres)
{
	struct file *file = cachefiles_cres_file(cres);

	if (file)
		fput(file);
	fscache_end_cookie_access(fscache_cres_cookie(cres), fscache_access_io_end);
}

static const struct netfs_cache_ops cachefiles_netfs_cache_ops = {
	.end_operation		= cachefiles_end_operation,
	.read			= cachefiles_read,
	.write			= cachefiles_write,
	.prepare_read		= cachefiles_prepare_read,
	.prepare_write		= cachefiles_prepare_write,
	.query_occupancy	= cachefiles_query_occupancy,
};

/*
 * Open the cache file when beginning a cache operation.
 */
bool cachefiles_begin_operation(struct netfs_cache_resources *cres,
				enum fscache_want_state want_state)
{
	struct cachefiles_object *object = cachefiles_cres_object(cres);

	if (!cachefiles_cres_file(cres)) {
		cres->ops = &cachefiles_netfs_cache_ops;
		if (object->file) {
			spin_lock(&object->lock);
			if (!cres->cache_priv2 && object->file)
				cres->cache_priv2 = get_file(object->file);
			spin_unlock(&object->lock);
		}
	}

	if (!cachefiles_cres_file(cres) && want_state != FSCACHE_WANT_PARAMS) {
		pr_err("failed to get cres->file\n");
		return false;
	}

	return true;
}
Commit	Line	Data
287fd611 DH	1	// SPDX-License-Identifier: GPL-2.0-or-later
	2	/* kiocb-using read/write
	3	*
	4	* Copyright (C) 2021 Red Hat, Inc. All Rights Reserved.
	5	* Written by David Howells (dhowells@redhat.com)
	6	*/
	7
	8	#include <linux/mount.h>
	9	#include <linux/slab.h>
	10	#include <linux/file.h>
	11	#include <linux/uio.h>
	12	#include <linux/falloc.h>
	13	#include <linux/sched/mm.h>
	14	#include <trace/events/fscache.h>
	15	#include "internal.h"
	16
047487c9 DH	17	struct cachefiles_kiocb {
	18	struct kiocb iocb;
	19	refcount_t ki_refcnt;
	20	loff_t start;
	21	union {
	22	size_t skipped;
	23	size_t len;
	24	};
	25	struct cachefiles_object *object;
	26	netfs_io_terminated_t term_func;
	27	void *term_func_priv;
	28	bool was_async;
	29	unsigned int inval_counter; /* Copy of cookie->inval_counter */
	30	u64 b_writing;
	31	};
	32
	33	static inline void cachefiles_put_kiocb(struct cachefiles_kiocb *ki)
	34	{
	35	if (refcount_dec_and_test(&ki->ki_refcnt)) {
	36	cachefiles_put_object(ki->object, cachefiles_obj_put_ioreq);
	37	fput(ki->iocb.ki_filp);
	38	kfree(ki);
	39	}
	40	}
	41
	42	/*
	43	* Handle completion of a read from the cache.
	44	*/
	45	static void cachefiles_read_complete(struct kiocb *iocb, long ret)
	46	{
	47	struct cachefiles_kiocb *ki = container_of(iocb, struct cachefiles_kiocb, iocb);
	48	struct inode *inode = file_inode(ki->iocb.ki_filp);
	49
	50	_enter("%ld", ret);
	51
	52	if (ret < 0)
	53	trace_cachefiles_io_error(ki->object, inode, ret,
	54	cachefiles_trace_read_error);
	55
	56	if (ki->term_func) {
	57	if (ret >= 0) {
	58	if (ki->object->cookie->inval_counter == ki->inval_counter)
	59	ki->skipped += ret;
	60	else
	61	ret = -ESTALE;
	62	}
	63
	64	ki->term_func(ki->term_func_priv, ret, ki->was_async);
	65	}
	66
	67	cachefiles_put_kiocb(ki);
	68	}
	69
	70	/*
	71	* Initiate a read from the cache.
	72	*/
	73	static int cachefiles_read(struct netfs_cache_resources *cres,
	74	loff_t start_pos,
	75	struct iov_iter *iter,
	76	enum netfs_read_from_hole read_hole,
	77	netfs_io_terminated_t term_func,
	78	void *term_func_priv)
	79	{
	80	struct cachefiles_object *object;
81	struct cachefiles_kiocb *ki;
82	struct file *file;
83	unsigned int old_nofs;
84	ssize_t ret = -ENOBUFS;
85	size_t len = iov_iter_count(iter), skipped = 0;
86
87	if (!fscache_wait_for_operation(cres, FSCACHE_WANT_READ))
88	goto presubmission_error;
89
90	fscache_count_read();
91	object = cachefiles_cres_object(cres);
92	file = cachefiles_cres_file(cres);
93
94	_enter("%pD,%li,%llx,%zx/%llx",
95	file, file_inode(file)->i_ino, start_pos, len,
96	i_size_read(file_inode(file)));
97
98	/* If the caller asked us to seek for data before doing the read, then
99	* we should do that now. If we find a gap, we fill it with zeros.
100	*/
101	if (read_hole != NETFS_READ_HOLE_IGNORE) {
102	loff_t off = start_pos, off2;
103
104	off2 = cachefiles_inject_read_error();
105	if (off2 == 0)
106	off2 = vfs_llseek(file, off, SEEK_DATA);
107	if (off2 < 0 && off2 >= (loff_t)-MAX_ERRNO && off2 != -ENXIO) {
108	skipped = 0;
109	ret = off2;
110	goto presubmission_error;
111	}
112
113	if (off2 == -ENXIO \|\| off2 >= start_pos + len) {
114	/* The region is beyond the EOF or there's no more data
115	* in the region, so clear the rest of the buffer and
116	* return success.
117	*/
118	ret = -ENODATA;
119	if (read_hole == NETFS_READ_HOLE_FAIL)
120	goto presubmission_error;
121
122	iov_iter_zero(len, iter);
123	skipped = len;
124	ret = 0;
125	goto presubmission_error;
126	}
127
128	skipped = off2 - off;
129	iov_iter_zero(skipped, iter);
130	}
131
132	ret = -ENOMEM;
133	ki = kzalloc(sizeof(struct cachefiles_kiocb), GFP_KERNEL);
134	if (!ki)
135	goto presubmission_error;
136
137	refcount_set(&ki->ki_refcnt, 2);
138	ki->iocb.ki_filp = file;
139	ki->iocb.ki_pos = start_pos + skipped;
140	ki->iocb.ki_flags = IOCB_DIRECT;
141	ki->iocb.ki_hint = ki_hint_validate(file_write_hint(file));
142	ki->iocb.ki_ioprio = get_current_ioprio();
143	ki->skipped = skipped;
144	ki->object = object;
145	ki->inval_counter = cres->inval_counter;
146	ki->term_func = term_func;
147	ki->term_func_priv = term_func_priv;
148	ki->was_async = true;
149
150	if (ki->term_func)
151	ki->iocb.ki_complete = cachefiles_read_complete;
152
153	get_file(ki->iocb.ki_filp);
154	cachefiles_grab_object(object, cachefiles_obj_get_ioreq);
155
156	trace_cachefiles_read(object, file_inode(file), ki->iocb.ki_pos, len - skipped);
157	old_nofs = memalloc_nofs_save();
158	ret = cachefiles_inject_read_error();
159	if (ret == 0)
160	ret = vfs_iocb_iter_read(file, &ki->iocb, iter);
161	memalloc_nofs_restore(old_nofs);
162	switch (ret) {
163	case -EIOCBQUEUED:
164	goto in_progress;
165
166	case -ERESTARTSYS:
167	case -ERESTARTNOINTR:
168	case -ERESTARTNOHAND:
169	case -ERESTART_RESTARTBLOCK:
170	/* There's no easy way to restart the syscall since other AIO's
171	* may be already running. Just fail this IO with EINTR.
172	*/
173	ret = -EINTR;
174	fallthrough;
175	default:
176	ki->was_async = false;
177	cachefiles_read_complete(&ki->iocb, ret);
178	if (ret > 0)
179	ret = 0;
180	break;
181	}
182
183	in_progress:
184	cachefiles_put_kiocb(ki);
185	_leave(" = %zd", ret);
186	return ret;
187
188	presubmission_error:
189	if (term_func)
190	term_func(term_func_priv, ret < 0 ? ret : skipped, false);
191	return ret;
192	}
193
bee9f655 DH	194	/*
	195	* Query the occupancy of the cache in a region, returning where the next chunk
	196	* of data starts and how long it is.
	197	*/
	198	static int cachefiles_query_occupancy(struct netfs_cache_resources *cres,
	199	loff_t start, size_t len, size_t granularity,
	200	loff_t _data_start, size_t _data_len)
	201	{
	202	struct cachefiles_object *object;
	203	struct file *file;
	204	loff_t off, off2;
	205
	206	*_data_start = -1;
	207	*_data_len = 0;
	208
	209	if (!fscache_wait_for_operation(cres, FSCACHE_WANT_READ))
	210	return -ENOBUFS;
	211
	212	object = cachefiles_cres_object(cres);
	213	file = cachefiles_cres_file(cres);
	214	granularity = max_t(size_t, object->volume->cache->bsize, granularity);
	215
	216	_enter("%pD,%li,%llx,%zx/%llx",
	217	file, file_inode(file)->i_ino, start, len,
	218	i_size_read(file_inode(file)));
	219
	220	off = cachefiles_inject_read_error();
	221	if (off == 0)
	222	off = vfs_llseek(file, start, SEEK_DATA);
	223	if (off == -ENXIO)
	224	return -ENODATA; /* Beyond EOF */
	225	if (off < 0 && off >= (loff_t)-MAX_ERRNO)
	226	return -ENOBUFS; /* Error. */
	227	if (round_up(off, granularity) >= start + len)
	228	return -ENODATA; /* No data in range */
	229
	230	off2 = cachefiles_inject_read_error();
	231	if (off2 == 0)
	232	off2 = vfs_llseek(file, off, SEEK_HOLE);
	233	if (off2 == -ENXIO)
	234	return -ENODATA; /* Beyond EOF */
	235	if (off2 < 0 && off2 >= (loff_t)-MAX_ERRNO)
	236	return -ENOBUFS; /* Error. */
	237
	238	/* Round away partial blocks */
	239	off = round_up(off, granularity);
	240	off2 = round_down(off2, granularity);
	241	if (off2 <= off)
	242	return -ENODATA;
	243
	244	*_data_start = off;
	245	if (off2 > start + len)
	246	*_data_len = len;
	247	else
	248	*_data_len = off2 - off;
	249	return 0;
	250	}
	251
047487c9 DH	252	/*
	253	* Handle completion of a write to the cache.
	254	*/
	255	static void cachefiles_write_complete(struct kiocb *iocb, long ret)
	256	{
	257	struct cachefiles_kiocb *ki = container_of(iocb, struct cachefiles_kiocb, iocb);
	258	struct cachefiles_object *object = ki->object;
	259	struct inode *inode = file_inode(ki->iocb.ki_filp);
	260
	261	_enter("%ld", ret);
	262
	263	/* Tell lockdep we inherited freeze protection from submission thread */
	264	__sb_writers_acquired(inode->i_sb, SB_FREEZE_WRITE);
	265	__sb_end_write(inode->i_sb, SB_FREEZE_WRITE);
	266
	267	if (ret < 0)
	268	trace_cachefiles_io_error(object, inode, ret,
	269	cachefiles_trace_write_error);
	270
	271	atomic_long_sub(ki->b_writing, &object->volume->cache->b_writing);
	272	set_bit(FSCACHE_COOKIE_HAVE_DATA, &object->cookie->flags);
	273	if (ki->term_func)
	274	ki->term_func(ki->term_func_priv, ret, ki->was_async);
	275	cachefiles_put_kiocb(ki);
	276	}
	277
	278	/*
	279	* Initiate a write to the cache.
	280	*/
	281	static int cachefiles_write(struct netfs_cache_resources *cres,
	282	loff_t start_pos,
	283	struct iov_iter *iter,
	284	netfs_io_terminated_t term_func,
	285	void *term_func_priv)
	286	{
	287	struct cachefiles_object *object;
	288	struct cachefiles_cache *cache;
	289	struct cachefiles_kiocb *ki;
	290	struct inode *inode;
	291	struct file *file;
	292	unsigned int old_nofs;
	293	ssize_t ret = -ENOBUFS;
	294	size_t len = iov_iter_count(iter);
	295
	296	if (!fscache_wait_for_operation(cres, FSCACHE_WANT_WRITE))
	297	goto presubmission_error;
	298	fscache_count_write();
	299	object = cachefiles_cres_object(cres);
	300	cache = object->volume->cache;
	301	file = cachefiles_cres_file(cres);
	302
	303	_enter("%pD,%li,%llx,%zx/%llx",
	304	file, file_inode(file)->i_ino, start_pos, len,
	305	i_size_read(file_inode(file)));
	306
	307	ret = -ENOMEM;
	308	ki = kzalloc(sizeof(struct cachefiles_kiocb), GFP_KERNEL);
	309	if (!ki)
	310	goto presubmission_error;
	311
	312	refcount_set(&ki->ki_refcnt, 2);
	313	ki->iocb.ki_filp = file;
	314	ki->iocb.ki_pos = start_pos;
	315	ki->iocb.ki_flags = IOCB_DIRECT \| IOCB_WRITE;
316	ki->iocb.ki_hint = ki_hint_validate(file_write_hint(file));
317	ki->iocb.ki_ioprio = get_current_ioprio();
318	ki->object = object;
319	ki->inval_counter = cres->inval_counter;
320	ki->start = start_pos;
321	ki->len = len;
322	ki->term_func = term_func;
323	ki->term_func_priv = term_func_priv;
324	ki->was_async = true;
5638b067	325	ki->b_writing = (len + (1 << cache->bshift) - 1) >> cache->bshift;
047487c9 DH	326
	327	if (ki->term_func)
	328	ki->iocb.ki_complete = cachefiles_write_complete;
	329	atomic_long_add(ki->b_writing, &cache->b_writing);
	330
	331	/* Open-code file_start_write here to grab freeze protection, which
	332	* will be released by another thread in aio_complete_rw(). Fool
	333	* lockdep by telling it the lock got released so that it doesn't
	334	* complain about the held lock when we return to userspace.
	335	*/
	336	inode = file_inode(file);
	337	__sb_start_write(inode->i_sb, SB_FREEZE_WRITE);
	338	__sb_writers_release(inode->i_sb, SB_FREEZE_WRITE);
	339
	340	get_file(ki->iocb.ki_filp);
	341	cachefiles_grab_object(object, cachefiles_obj_get_ioreq);
	342
	343	trace_cachefiles_write(object, inode, ki->iocb.ki_pos, len);
	344	old_nofs = memalloc_nofs_save();
	345	ret = cachefiles_inject_write_error();
	346	if (ret == 0)
	347	ret = vfs_iocb_iter_write(file, &ki->iocb, iter);
	348	memalloc_nofs_restore(old_nofs);
	349	switch (ret) {
	350	case -EIOCBQUEUED:
	351	goto in_progress;
	352
	353	case -ERESTARTSYS:
	354	case -ERESTARTNOINTR:
	355	case -ERESTARTNOHAND:
	356	case -ERESTART_RESTARTBLOCK:
	357	/* There's no easy way to restart the syscall since other AIO's
	358	* may be already running. Just fail this IO with EINTR.
	359	*/
	360	ret = -EINTR;
	361	fallthrough;
	362	default:
	363	ki->was_async = false;
	364	cachefiles_write_complete(&ki->iocb, ret);
	365	if (ret > 0)
	366	ret = 0;
	367	break;
	368	}
	369
	370	in_progress:
	371	cachefiles_put_kiocb(ki);
	372	_leave(" = %zd", ret);
	373	return ret;
	374
	375	presubmission_error:
	376	if (term_func)
	377	term_func(term_func_priv, ret, false);
	378	return ret;
	379	}
	380
	381	/*
	382	* Prepare a read operation, shortening it to a cached/uncached
	383	* boundary as appropriate.
	384	*/
6a19114b	385	static enum netfs_io_source cachefiles_prepare_read(struct netfs_io_subrequest *subreq,
047487c9 DH	386	loff_t i_size)
	387	{
	388	enum cachefiles_prepare_read_trace why;
6a19114b	389	struct netfs_io_request *rreq = subreq->rreq;
047487c9 DH	390	struct netfs_cache_resources *cres = &rreq->cache_resources;
	391	struct cachefiles_object *object;
	392	struct cachefiles_cache *cache;
	393	struct fscache_cookie *cookie = fscache_cres_cookie(cres);
	394	const struct cred *saved_cred;
	395	struct file *file = cachefiles_cres_file(cres);
6a19114b	396	enum netfs_io_source ret = NETFS_DOWNLOAD_FROM_SERVER;
047487c9 DH	397	loff_t off, to;
	398	ino_t ino = file ? file_inode(file)->i_ino : 0;
	399
	400	_enter("%zx @%llx/%llx", subreq->len, subreq->start, i_size);
	401
	402	if (subreq->start >= i_size) {
	403	ret = NETFS_FILL_WITH_ZEROES;
	404	why = cachefiles_trace_read_after_eof;
	405	goto out_no_object;
	406	}
	407
	408	if (test_bit(FSCACHE_COOKIE_NO_DATA_TO_READ, &cookie->flags)) {
	409	__set_bit(NETFS_SREQ_WRITE_TO_CACHE, &subreq->flags);
	410	why = cachefiles_trace_read_no_data;
	411	goto out_no_object;
	412	}
	413
	414	/* The object and the file may be being created in the background. */
	415	if (!file) {
	416	why = cachefiles_trace_read_no_file;
	417	if (!fscache_wait_for_operation(cres, FSCACHE_WANT_READ))
	418	goto out_no_object;
	419	file = cachefiles_cres_file(cres);
	420	if (!file)
	421	goto out_no_object;
	422	ino = file_inode(file)->i_ino;
	423	}
	424
	425	object = cachefiles_cres_object(cres);
	426	cache = object->volume->cache;
	427	cachefiles_begin_secure(cache, &saved_cred);
	428
	429	off = cachefiles_inject_read_error();
	430	if (off == 0)
	431	off = vfs_llseek(file, subreq->start, SEEK_DATA);
	432	if (off < 0 && off >= (loff_t)-MAX_ERRNO) {
	433	if (off == (loff_t)-ENXIO) {
	434	why = cachefiles_trace_read_seek_nxio;
	435	goto download_and_store;
	436	}
	437	trace_cachefiles_io_error(object, file_inode(file), off,
	438	cachefiles_trace_seek_error);
	439	why = cachefiles_trace_read_seek_error;
	440	goto out;
	441	}
	442
	443	if (off >= subreq->start + subreq->len) {
	444	why = cachefiles_trace_read_found_hole;
	445	goto download_and_store;
	446	}
	447
	448	if (off > subreq->start) {
	449	off = round_up(off, cache->bsize);
	450	subreq->len = off - subreq->start;
	451	why = cachefiles_trace_read_found_part;
	452	goto download_and_store;
	453	}
	454
	455	to = cachefiles_inject_read_error();
	456	if (to == 0)
	457	to = vfs_llseek(file, subreq->start, SEEK_HOLE);
	458	if (to < 0 && to >= (loff_t)-MAX_ERRNO) {
	459	trace_cachefiles_io_error(object, file_inode(file), to,
	460	cachefiles_trace_seek_error);
461	why = cachefiles_trace_read_seek_error;
462	goto out;
463	}
464
465	if (to < subreq->start + subreq->len) {
466	if (subreq->start + subreq->len >= i_size)
467	to = round_up(to, cache->bsize);
468	else
469	to = round_down(to, cache->bsize);
470	subreq->len = to - subreq->start;
471	}
472
473	why = cachefiles_trace_read_have_data;
474	ret = NETFS_READ_FROM_CACHE;
475	goto out;
476
477	download_and_store:
478	__set_bit(NETFS_SREQ_WRITE_TO_CACHE, &subreq->flags);
479	out:
480	cachefiles_end_secure(cache, saved_cred);
481	out_no_object:
482	trace_cachefiles_prep_read(subreq, ret, why, ino);
483	return ret;
484	}
485
486	/*
487	* Prepare for a write to occur.
488	*/
489	static int __cachefiles_prepare_write(struct netfs_cache_resources *cres,
490	loff_t _start, size_t _len, loff_t i_size,
491	bool no_space_allocated_yet)
492	{
493	struct cachefiles_object *object = cachefiles_cres_object(cres);
494	struct cachefiles_cache *cache = object->volume->cache;
495	struct file *file = cachefiles_cres_file(cres);
496	loff_t start = *_start, pos;
497	size_t len = *_len, down;
498	int ret;
499
500	/* Round to DIO size */
501	down = start - round_down(start, PAGE_SIZE);
502	*_start = start - down;
503	*_len = round_up(down + len, PAGE_SIZE);
504
505	/* We need to work out whether there's sufficient disk space to perform
506	* the write - but we can skip that check if we have space already
507	* allocated.
508	*/
509	if (no_space_allocated_yet)
510	goto check_space;
511
512	pos = cachefiles_inject_read_error();
513	if (pos == 0)
514	pos = vfs_llseek(file, *_start, SEEK_DATA);
515	if (pos < 0 && pos >= (loff_t)-MAX_ERRNO) {
516	if (pos == -ENXIO)
517	goto check_space; /* Unallocated tail */
518	trace_cachefiles_io_error(object, file_inode(file), pos,
519	cachefiles_trace_seek_error);
520	return pos;
521	}
522	if ((u64)pos >= (u64)_start + _len)
523	goto check_space; /* Unallocated region */
524
525	/* We have a block that's at least partially filled - if we're low on
526	* space, we need to see if it's fully allocated. If it's not, we may
527	* want to cull it.
528	*/
3929eca7 DH	529	if (cachefiles_has_space(cache, 0, *_len / PAGE_SIZE,
3929eca7 DH	530	cachefiles_has_space_check) == 0)
047487c9 DH	531	return 0; /* Enough space to simply overwrite the whole block */
	532
	533	pos = cachefiles_inject_read_error();
	534	if (pos == 0)
	535	pos = vfs_llseek(file, *_start, SEEK_HOLE);
	536	if (pos < 0 && pos >= (loff_t)-MAX_ERRNO) {
	537	trace_cachefiles_io_error(object, file_inode(file), pos,
	538	cachefiles_trace_seek_error);
	539	return pos;
	540	}
	541	if ((u64)pos >= (u64)_start + _len)
	542	return 0; /* Fully allocated */
	543
	544	/* Partially allocated, but insufficient space: cull. */
3929eca7	545	fscache_count_no_write_space();
047487c9 DH	546	ret = cachefiles_inject_remove_error();
	547	if (ret == 0)
	548	ret = vfs_fallocate(file, FALLOC_FL_PUNCH_HOLE \| FALLOC_FL_KEEP_SIZE,
	549	_start, _len);
	550	if (ret < 0) {
	551	trace_cachefiles_io_error(object, file_inode(file), ret,
	552	cachefiles_trace_fallocate_error);
	553	cachefiles_io_error_obj(object,
	554	"CacheFiles: fallocate failed (%d)\n", ret);
	555	ret = -EIO;
	556	}
	557
	558	return ret;
	559
	560	check_space:
3929eca7 DH	561	return cachefiles_has_space(cache, 0, *_len / PAGE_SIZE,
3929eca7 DH	562	cachefiles_has_space_for_write);
047487c9 DH	563	}
	564
	565	static int cachefiles_prepare_write(struct netfs_cache_resources *cres,
	566	loff_t _start, size_t _len, loff_t i_size,
	567	bool no_space_allocated_yet)
	568	{
	569	struct cachefiles_object *object = cachefiles_cres_object(cres);
	570	struct cachefiles_cache *cache = object->volume->cache;
	571	const struct cred *saved_cred;
	572	int ret;
	573
	574	if (!cachefiles_cres_file(cres)) {
	575	if (!fscache_wait_for_operation(cres, FSCACHE_WANT_WRITE))
	576	return -ENOBUFS;
	577	if (!cachefiles_cres_file(cres))
	578	return -ENOBUFS;
	579	}
	580
	581	cachefiles_begin_secure(cache, &saved_cred);
	582	ret = __cachefiles_prepare_write(cres, _start, _len, i_size,
	583	no_space_allocated_yet);
	584	cachefiles_end_secure(cache, saved_cred);
	585	return ret;
	586	}
	587
287fd611 DH	588	/*
	589	* Clean up an operation.
	590	*/
	591	static void cachefiles_end_operation(struct netfs_cache_resources *cres)
	592	{
	593	struct file *file = cachefiles_cres_file(cres);
	594
	595	if (file)
	596	fput(file);
	597	fscache_end_cookie_access(fscache_cres_cookie(cres), fscache_access_io_end);
	598	}
	599
	600	static const struct netfs_cache_ops cachefiles_netfs_cache_ops = {
	601	.end_operation = cachefiles_end_operation,
047487c9 DH	602	.read = cachefiles_read,
	603	.write = cachefiles_write,
	604	.prepare_read = cachefiles_prepare_read,
	605	.prepare_write = cachefiles_prepare_write,
bee9f655	606	.query_occupancy = cachefiles_query_occupancy,
287fd611 DH	607	};
	608
	609	/*
	610	* Open the cache file when beginning a cache operation.
	611	*/
	612	bool cachefiles_begin_operation(struct netfs_cache_resources *cres,
	613	enum fscache_want_state want_state)
	614	{
	615	struct cachefiles_object *object = cachefiles_cres_object(cres);
	616
	617	if (!cachefiles_cres_file(cres)) {
	618	cres->ops = &cachefiles_netfs_cache_ops;
	619	if (object->file) {
	620	spin_lock(&object->lock);
	621	if (!cres->cache_priv2 && object->file)
	622	cres->cache_priv2 = get_file(object->file);
	623	spin_unlock(&object->lock);
	624	}
	625	}
	626
	627	if (!cachefiles_cres_file(cres) && want_state != FSCACHE_WANT_PARAMS) {
	628	pr_err("failed to get cres->file\n");
	629	return false;
	630	}
	631
	632	return true;
	633	}