linux/fs/netfs/iterator.c
David Howells ee4cdf7ba8
netfs: Speed up buffered reading
Improve the efficiency of buffered reads in a number of ways:

 (1) Overhaul the algorithm in general so that it's a lot more compact and
     split the read submission code between buffered and unbuffered
     versions.  The unbuffered version can be vastly simplified.

 (2) Read-result collection is handed off to a work queue rather than being
     done in the I/O thread.  Multiple subrequests can be processes
     simultaneously.

 (3) When a subrequest is collected, any folios it fully spans are
     collected and "spare" data on either side is donated to either the
     previous or the next subrequest in the sequence.

Notes:

 (*) Readahead expansion is massively slows down fio, presumably because it
     causes a load of extra allocations, both folio and xarray, up front
     before RPC requests can be transmitted.

 (*) RDMA with cifs does appear to work, both with SIW and RXE.

 (*) PG_private_2-based reading and copy-to-cache is split out into its own
     file and altered to use folio_queue.  Note that the copy to the cache
     now creates a new write transaction against the cache and adds the
     folios to be copied into it.  This allows it to use part of the
     writeback I/O code.

Signed-off-by: David Howells <dhowells@redhat.com>
cc: Jeff Layton <jlayton@kernel.org>
cc: netfs@lists.linux.dev
cc: linux-fsdevel@vger.kernel.org
Link: https://lore.kernel.org/r/20240814203850.2240469-20-dhowells@redhat.com/ # v2
Signed-off-by: Christian Brauner <brauner@kernel.org>
2024-09-12 12:20:41 +02:00

251 lines
6.5 KiB
C

// SPDX-License-Identifier: GPL-2.0-or-later
/* Iterator helpers.
*
* Copyright (C) 2022 Red Hat, Inc. All Rights Reserved.
* Written by David Howells (dhowells@redhat.com)
*/
#include <linux/export.h>
#include <linux/slab.h>
#include <linux/mm.h>
#include <linux/uio.h>
#include <linux/scatterlist.h>
#include <linux/netfs.h>
#include "internal.h"
/**
* netfs_extract_user_iter - Extract the pages from a user iterator into a bvec
* @orig: The original iterator
* @orig_len: The amount of iterator to copy
* @new: The iterator to be set up
* @extraction_flags: Flags to qualify the request
*
* Extract the page fragments from the given amount of the source iterator and
* build up a second iterator that refers to all of those bits. This allows
* the original iterator to disposed of.
*
* @extraction_flags can have ITER_ALLOW_P2PDMA set to request peer-to-peer DMA be
* allowed on the pages extracted.
*
* On success, the number of elements in the bvec is returned, the original
* iterator will have been advanced by the amount extracted.
*
* The iov_iter_extract_mode() function should be used to query how cleanup
* should be performed.
*/
ssize_t netfs_extract_user_iter(struct iov_iter *orig, size_t orig_len,
struct iov_iter *new,
iov_iter_extraction_t extraction_flags)
{
struct bio_vec *bv = NULL;
struct page **pages;
unsigned int cur_npages;
unsigned int max_pages;
unsigned int npages = 0;
unsigned int i;
ssize_t ret;
size_t count = orig_len, offset, len;
size_t bv_size, pg_size;
if (WARN_ON_ONCE(!iter_is_ubuf(orig) && !iter_is_iovec(orig)))
return -EIO;
max_pages = iov_iter_npages(orig, INT_MAX);
bv_size = array_size(max_pages, sizeof(*bv));
bv = kvmalloc(bv_size, GFP_KERNEL);
if (!bv)
return -ENOMEM;
/* Put the page list at the end of the bvec list storage. bvec
* elements are larger than page pointers, so as long as we work
* 0->last, we should be fine.
*/
pg_size = array_size(max_pages, sizeof(*pages));
pages = (void *)bv + bv_size - pg_size;
while (count && npages < max_pages) {
ret = iov_iter_extract_pages(orig, &pages, count,
max_pages - npages, extraction_flags,
&offset);
if (ret < 0) {
pr_err("Couldn't get user pages (rc=%zd)\n", ret);
break;
}
if (ret > count) {
pr_err("get_pages rc=%zd more than %zu\n", ret, count);
break;
}
count -= ret;
ret += offset;
cur_npages = DIV_ROUND_UP(ret, PAGE_SIZE);
if (npages + cur_npages > max_pages) {
pr_err("Out of bvec array capacity (%u vs %u)\n",
npages + cur_npages, max_pages);
break;
}
for (i = 0; i < cur_npages; i++) {
len = ret > PAGE_SIZE ? PAGE_SIZE : ret;
bvec_set_page(bv + npages + i, *pages++, len - offset, offset);
ret -= len;
offset = 0;
}
npages += cur_npages;
}
iov_iter_bvec(new, orig->data_source, bv, npages, orig_len - count);
return npages;
}
EXPORT_SYMBOL_GPL(netfs_extract_user_iter);
/*
* Select the span of a bvec iterator we're going to use. Limit it by both maximum
* size and maximum number of segments. Returns the size of the span in bytes.
*/
static size_t netfs_limit_bvec(const struct iov_iter *iter, size_t start_offset,
size_t max_size, size_t max_segs)
{
const struct bio_vec *bvecs = iter->bvec;
unsigned int nbv = iter->nr_segs, ix = 0, nsegs = 0;
size_t len, span = 0, n = iter->count;
size_t skip = iter->iov_offset + start_offset;
if (WARN_ON(!iov_iter_is_bvec(iter)) ||
WARN_ON(start_offset > n) ||
n == 0)
return 0;
while (n && ix < nbv && skip) {
len = bvecs[ix].bv_len;
if (skip < len)
break;
skip -= len;
n -= len;
ix++;
}
while (n && ix < nbv) {
len = min3(n, bvecs[ix].bv_len - skip, max_size);
span += len;
nsegs++;
ix++;
if (span >= max_size || nsegs >= max_segs)
break;
skip = 0;
n -= len;
}
return min(span, max_size);
}
/*
* Select the span of an xarray iterator we're going to use. Limit it by both
* maximum size and maximum number of segments. It is assumed that segments
* can be larger than a page in size, provided they're physically contiguous.
* Returns the size of the span in bytes.
*/
static size_t netfs_limit_xarray(const struct iov_iter *iter, size_t start_offset,
size_t max_size, size_t max_segs)
{
struct folio *folio;
unsigned int nsegs = 0;
loff_t pos = iter->xarray_start + iter->iov_offset;
pgoff_t index = pos / PAGE_SIZE;
size_t span = 0, n = iter->count;
XA_STATE(xas, iter->xarray, index);
if (WARN_ON(!iov_iter_is_xarray(iter)) ||
WARN_ON(start_offset > n) ||
n == 0)
return 0;
max_size = min(max_size, n - start_offset);
rcu_read_lock();
xas_for_each(&xas, folio, ULONG_MAX) {
size_t offset, flen, len;
if (xas_retry(&xas, folio))
continue;
if (WARN_ON(xa_is_value(folio)))
break;
if (WARN_ON(folio_test_hugetlb(folio)))
break;
flen = folio_size(folio);
offset = offset_in_folio(folio, pos);
len = min(max_size, flen - offset);
span += len;
nsegs++;
if (span >= max_size || nsegs >= max_segs)
break;
}
rcu_read_unlock();
return min(span, max_size);
}
/*
* Select the span of a folio queue iterator we're going to use. Limit it by
* both maximum size and maximum number of segments. Returns the size of the
* span in bytes.
*/
static size_t netfs_limit_folioq(const struct iov_iter *iter, size_t start_offset,
size_t max_size, size_t max_segs)
{
const struct folio_queue *folioq = iter->folioq;
unsigned int nsegs = 0;
unsigned int slot = iter->folioq_slot;
size_t span = 0, n = iter->count;
if (WARN_ON(!iov_iter_is_folioq(iter)) ||
WARN_ON(start_offset > n) ||
n == 0)
return 0;
max_size = umin(max_size, n - start_offset);
if (slot >= folioq_nr_slots(folioq)) {
folioq = folioq->next;
slot = 0;
}
start_offset += iter->iov_offset;
do {
size_t flen = folioq_folio_size(folioq, slot);
if (start_offset < flen) {
span += flen - start_offset;
nsegs++;
start_offset = 0;
} else {
start_offset -= flen;
}
if (span >= max_size || nsegs >= max_segs)
break;
slot++;
if (slot >= folioq_nr_slots(folioq)) {
folioq = folioq->next;
slot = 0;
}
} while (folioq);
return umin(span, max_size);
}
size_t netfs_limit_iter(const struct iov_iter *iter, size_t start_offset,
size_t max_size, size_t max_segs)
{
if (iov_iter_is_folioq(iter))
return netfs_limit_folioq(iter, start_offset, max_size, max_segs);
if (iov_iter_is_bvec(iter))
return netfs_limit_bvec(iter, start_offset, max_size, max_segs);
if (iov_iter_is_xarray(iter))
return netfs_limit_xarray(iter, start_offset, max_size, max_segs);
BUG();
}
EXPORT_SYMBOL(netfs_limit_iter);