2019-07-23 00:26:22 +08:00
|
|
|
// SPDX-License-Identifier: GPL-2.0
|
|
|
|
/*
|
2020-11-14 05:19:15 +08:00
|
|
|
* Data verification functions, i.e. hooks for ->readpages()
|
2019-07-23 00:26:22 +08:00
|
|
|
*
|
|
|
|
* Copyright 2019 Google LLC
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include "fsverity_private.h"
|
|
|
|
|
|
|
|
#include <crypto/hash.h>
|
|
|
|
#include <linux/bio.h>
|
|
|
|
#include <linux/ratelimit.h>
|
|
|
|
|
|
|
|
static struct workqueue_struct *fsverity_read_workqueue;
|
|
|
|
|
|
|
|
/**
|
|
|
|
* hash_at_level() - compute the location of the block's hash at the given level
|
|
|
|
*
|
|
|
|
* @params: (in) the Merkle tree parameters
|
|
|
|
* @dindex: (in) the index of the data block being verified
|
|
|
|
* @level: (in) the level of hash we want (0 is leaf level)
|
|
|
|
* @hindex: (out) the index of the hash block containing the wanted hash
|
|
|
|
* @hoffset: (out) the byte offset to the wanted hash within the hash block
|
|
|
|
*/
|
|
|
|
static void hash_at_level(const struct merkle_tree_params *params,
|
|
|
|
pgoff_t dindex, unsigned int level, pgoff_t *hindex,
|
|
|
|
unsigned int *hoffset)
|
|
|
|
{
|
|
|
|
pgoff_t position;
|
|
|
|
|
|
|
|
/* Offset of the hash within the level's region, in hashes */
|
|
|
|
position = dindex >> (level * params->log_arity);
|
|
|
|
|
|
|
|
/* Index of the hash block in the tree overall */
|
|
|
|
*hindex = params->level_start[level] + (position >> params->log_arity);
|
|
|
|
|
|
|
|
/* Offset of the wanted hash (in bytes) within the hash block */
|
|
|
|
*hoffset = (position & ((1 << params->log_arity) - 1)) <<
|
|
|
|
(params->log_blocksize - params->log_arity);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Extract a hash from a hash page */
|
|
|
|
static void extract_hash(struct page *hpage, unsigned int hoffset,
|
|
|
|
unsigned int hsize, u8 *out)
|
|
|
|
{
|
|
|
|
void *virt = kmap_atomic(hpage);
|
|
|
|
|
|
|
|
memcpy(out, virt + hoffset, hsize);
|
|
|
|
kunmap_atomic(virt);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline int cmp_hashes(const struct fsverity_info *vi,
|
|
|
|
const u8 *want_hash, const u8 *real_hash,
|
|
|
|
pgoff_t index, int level)
|
|
|
|
{
|
|
|
|
const unsigned int hsize = vi->tree_params.digest_size;
|
|
|
|
|
|
|
|
if (memcmp(want_hash, real_hash, hsize) == 0)
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
fsverity_err(vi->inode,
|
|
|
|
"FILE CORRUPTED! index=%lu, level=%d, want_hash=%s:%*phN, real_hash=%s:%*phN",
|
|
|
|
index, level,
|
|
|
|
vi->tree_params.hash_alg->name, hsize, want_hash,
|
|
|
|
vi->tree_params.hash_alg->name, hsize, real_hash);
|
|
|
|
return -EBADMSG;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Verify a single data page against the file's Merkle tree.
|
|
|
|
*
|
|
|
|
* In principle, we need to verify the entire path to the root node. However,
|
|
|
|
* for efficiency the filesystem may cache the hash pages. Therefore we need
|
|
|
|
* only ascend the tree until an already-verified page is seen, as indicated by
|
|
|
|
* the PageChecked bit being set; then verify the path to that page.
|
|
|
|
*
|
|
|
|
* This code currently only supports the case where the verity block size is
|
|
|
|
* equal to PAGE_SIZE. Doing otherwise would be possible but tricky, since we
|
|
|
|
* wouldn't be able to use the PageChecked bit.
|
|
|
|
*
|
|
|
|
* Note that multiple processes may race to verify a hash page and mark it
|
|
|
|
* Checked, but it doesn't matter; the result will be the same either way.
|
|
|
|
*
|
|
|
|
* Return: true if the page is valid, else false.
|
|
|
|
*/
|
|
|
|
static bool verify_page(struct inode *inode, const struct fsverity_info *vi,
|
fs-verity: implement readahead of Merkle tree pages
When fs-verity verifies data pages, currently it reads each Merkle tree
page synchronously using read_mapping_page().
Therefore, when the Merkle tree pages aren't already cached, fs-verity
causes an extra 4 KiB I/O request for every 512 KiB of data (assuming
that the Merkle tree uses SHA-256 and 4 KiB blocks). This results in
more I/O requests and performance loss than is strictly necessary.
Therefore, implement readahead of the Merkle tree pages.
For simplicity, we take advantage of the fact that the kernel already
does readahead of the file's *data*, just like it does for any other
file. Due to this, we don't really need a separate readahead state
(struct file_ra_state) just for the Merkle tree, but rather we just need
to piggy-back on the existing data readahead requests.
We also only really need to bother with the first level of the Merkle
tree, since the usual fan-out factor is 128, so normally over 99% of
Merkle tree I/O requests are for the first level.
Therefore, make fsverity_verify_bio() enable readahead of the first
Merkle tree level, for up to 1/4 the number of pages in the bio, when it
sees that the REQ_RAHEAD flag is set on the bio. The readahead size is
then passed down to ->read_merkle_tree_page() for the filesystem to
(optionally) implement if it sees that the requested page is uncached.
While we're at it, also make build_merkle_tree_level() set the Merkle
tree readahead size, since it's easy to do there.
However, for now don't set the readahead size in fsverity_verify_page(),
since currently it's only used to verify holes on ext4 and f2fs, and it
would need parameters added to know how much to read ahead.
This patch significantly improves fs-verity sequential read performance.
Some quick benchmarks with 'cat'-ing a 250MB file after dropping caches:
On an ARM64 phone (using sha256-ce):
Before: 217 MB/s
After: 263 MB/s
(compare to sha256sum of non-verity file: 357 MB/s)
In an x86_64 VM (using sha256-avx2):
Before: 173 MB/s
After: 215 MB/s
(compare to sha256sum of non-verity file: 223 MB/s)
Link: https://lore.kernel.org/r/20200106205533.137005-1-ebiggers@kernel.org
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-01-07 04:55:33 +08:00
|
|
|
struct ahash_request *req, struct page *data_page,
|
|
|
|
unsigned long level0_ra_pages)
|
2019-07-23 00:26:22 +08:00
|
|
|
{
|
|
|
|
const struct merkle_tree_params *params = &vi->tree_params;
|
|
|
|
const unsigned int hsize = params->digest_size;
|
|
|
|
const pgoff_t index = data_page->index;
|
|
|
|
int level;
|
|
|
|
u8 _want_hash[FS_VERITY_MAX_DIGEST_SIZE];
|
|
|
|
const u8 *want_hash;
|
|
|
|
u8 real_hash[FS_VERITY_MAX_DIGEST_SIZE];
|
|
|
|
struct page *hpages[FS_VERITY_MAX_LEVELS];
|
|
|
|
unsigned int hoffsets[FS_VERITY_MAX_LEVELS];
|
|
|
|
int err;
|
|
|
|
|
|
|
|
if (WARN_ON_ONCE(!PageLocked(data_page) || PageUptodate(data_page)))
|
|
|
|
return false;
|
|
|
|
|
|
|
|
pr_debug_ratelimited("Verifying data page %lu...\n", index);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Starting at the leaf level, ascend the tree saving hash pages along
|
|
|
|
* the way until we find a verified hash page, indicated by PageChecked;
|
|
|
|
* or until we reach the root.
|
|
|
|
*/
|
|
|
|
for (level = 0; level < params->num_levels; level++) {
|
|
|
|
pgoff_t hindex;
|
|
|
|
unsigned int hoffset;
|
|
|
|
struct page *hpage;
|
|
|
|
|
|
|
|
hash_at_level(params, index, level, &hindex, &hoffset);
|
|
|
|
|
|
|
|
pr_debug_ratelimited("Level %d: hindex=%lu, hoffset=%u\n",
|
|
|
|
level, hindex, hoffset);
|
|
|
|
|
fs-verity: implement readahead of Merkle tree pages
When fs-verity verifies data pages, currently it reads each Merkle tree
page synchronously using read_mapping_page().
Therefore, when the Merkle tree pages aren't already cached, fs-verity
causes an extra 4 KiB I/O request for every 512 KiB of data (assuming
that the Merkle tree uses SHA-256 and 4 KiB blocks). This results in
more I/O requests and performance loss than is strictly necessary.
Therefore, implement readahead of the Merkle tree pages.
For simplicity, we take advantage of the fact that the kernel already
does readahead of the file's *data*, just like it does for any other
file. Due to this, we don't really need a separate readahead state
(struct file_ra_state) just for the Merkle tree, but rather we just need
to piggy-back on the existing data readahead requests.
We also only really need to bother with the first level of the Merkle
tree, since the usual fan-out factor is 128, so normally over 99% of
Merkle tree I/O requests are for the first level.
Therefore, make fsverity_verify_bio() enable readahead of the first
Merkle tree level, for up to 1/4 the number of pages in the bio, when it
sees that the REQ_RAHEAD flag is set on the bio. The readahead size is
then passed down to ->read_merkle_tree_page() for the filesystem to
(optionally) implement if it sees that the requested page is uncached.
While we're at it, also make build_merkle_tree_level() set the Merkle
tree readahead size, since it's easy to do there.
However, for now don't set the readahead size in fsverity_verify_page(),
since currently it's only used to verify holes on ext4 and f2fs, and it
would need parameters added to know how much to read ahead.
This patch significantly improves fs-verity sequential read performance.
Some quick benchmarks with 'cat'-ing a 250MB file after dropping caches:
On an ARM64 phone (using sha256-ce):
Before: 217 MB/s
After: 263 MB/s
(compare to sha256sum of non-verity file: 357 MB/s)
In an x86_64 VM (using sha256-avx2):
Before: 173 MB/s
After: 215 MB/s
(compare to sha256sum of non-verity file: 223 MB/s)
Link: https://lore.kernel.org/r/20200106205533.137005-1-ebiggers@kernel.org
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-01-07 04:55:33 +08:00
|
|
|
hpage = inode->i_sb->s_vop->read_merkle_tree_page(inode, hindex,
|
|
|
|
level == 0 ? level0_ra_pages : 0);
|
2019-07-23 00:26:22 +08:00
|
|
|
if (IS_ERR(hpage)) {
|
|
|
|
err = PTR_ERR(hpage);
|
|
|
|
fsverity_err(inode,
|
|
|
|
"Error %d reading Merkle tree page %lu",
|
|
|
|
err, hindex);
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (PageChecked(hpage)) {
|
|
|
|
extract_hash(hpage, hoffset, hsize, _want_hash);
|
|
|
|
want_hash = _want_hash;
|
|
|
|
put_page(hpage);
|
|
|
|
pr_debug_ratelimited("Hash page already checked, want %s:%*phN\n",
|
|
|
|
params->hash_alg->name,
|
|
|
|
hsize, want_hash);
|
|
|
|
goto descend;
|
|
|
|
}
|
|
|
|
pr_debug_ratelimited("Hash page not yet checked\n");
|
|
|
|
hpages[level] = hpage;
|
|
|
|
hoffsets[level] = hoffset;
|
|
|
|
}
|
|
|
|
|
|
|
|
want_hash = vi->root_hash;
|
|
|
|
pr_debug("Want root hash: %s:%*phN\n",
|
|
|
|
params->hash_alg->name, hsize, want_hash);
|
|
|
|
descend:
|
|
|
|
/* Descend the tree verifying hash pages */
|
|
|
|
for (; level > 0; level--) {
|
|
|
|
struct page *hpage = hpages[level - 1];
|
|
|
|
unsigned int hoffset = hoffsets[level - 1];
|
|
|
|
|
|
|
|
err = fsverity_hash_page(params, inode, req, hpage, real_hash);
|
|
|
|
if (err)
|
|
|
|
goto out;
|
|
|
|
err = cmp_hashes(vi, want_hash, real_hash, index, level - 1);
|
|
|
|
if (err)
|
|
|
|
goto out;
|
|
|
|
SetPageChecked(hpage);
|
|
|
|
extract_hash(hpage, hoffset, hsize, _want_hash);
|
|
|
|
want_hash = _want_hash;
|
|
|
|
put_page(hpage);
|
|
|
|
pr_debug("Verified hash page at level %d, now want %s:%*phN\n",
|
|
|
|
level - 1, params->hash_alg->name, hsize, want_hash);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Finally, verify the data page */
|
|
|
|
err = fsverity_hash_page(params, inode, req, data_page, real_hash);
|
|
|
|
if (err)
|
|
|
|
goto out;
|
|
|
|
err = cmp_hashes(vi, want_hash, real_hash, index, -1);
|
|
|
|
out:
|
|
|
|
for (; level > 0; level--)
|
|
|
|
put_page(hpages[level - 1]);
|
|
|
|
|
|
|
|
return err == 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* fsverity_verify_page() - verify a data page
|
2020-05-12 03:21:17 +08:00
|
|
|
* @page: the page to verity
|
2019-07-23 00:26:22 +08:00
|
|
|
*
|
|
|
|
* Verify a page that has just been read from a verity file. The page must be a
|
|
|
|
* pagecache page that is still locked and not yet uptodate.
|
|
|
|
*
|
|
|
|
* Return: true if the page is valid, else false.
|
|
|
|
*/
|
|
|
|
bool fsverity_verify_page(struct page *page)
|
|
|
|
{
|
|
|
|
struct inode *inode = page->mapping->host;
|
|
|
|
const struct fsverity_info *vi = inode->i_verity_info;
|
|
|
|
struct ahash_request *req;
|
|
|
|
bool valid;
|
|
|
|
|
2020-01-01 01:55:45 +08:00
|
|
|
/* This allocation never fails, since it's mempool-backed. */
|
|
|
|
req = fsverity_alloc_hash_request(vi->tree_params.hash_alg, GFP_NOFS);
|
2019-07-23 00:26:22 +08:00
|
|
|
|
fs-verity: implement readahead of Merkle tree pages
When fs-verity verifies data pages, currently it reads each Merkle tree
page synchronously using read_mapping_page().
Therefore, when the Merkle tree pages aren't already cached, fs-verity
causes an extra 4 KiB I/O request for every 512 KiB of data (assuming
that the Merkle tree uses SHA-256 and 4 KiB blocks). This results in
more I/O requests and performance loss than is strictly necessary.
Therefore, implement readahead of the Merkle tree pages.
For simplicity, we take advantage of the fact that the kernel already
does readahead of the file's *data*, just like it does for any other
file. Due to this, we don't really need a separate readahead state
(struct file_ra_state) just for the Merkle tree, but rather we just need
to piggy-back on the existing data readahead requests.
We also only really need to bother with the first level of the Merkle
tree, since the usual fan-out factor is 128, so normally over 99% of
Merkle tree I/O requests are for the first level.
Therefore, make fsverity_verify_bio() enable readahead of the first
Merkle tree level, for up to 1/4 the number of pages in the bio, when it
sees that the REQ_RAHEAD flag is set on the bio. The readahead size is
then passed down to ->read_merkle_tree_page() for the filesystem to
(optionally) implement if it sees that the requested page is uncached.
While we're at it, also make build_merkle_tree_level() set the Merkle
tree readahead size, since it's easy to do there.
However, for now don't set the readahead size in fsverity_verify_page(),
since currently it's only used to verify holes on ext4 and f2fs, and it
would need parameters added to know how much to read ahead.
This patch significantly improves fs-verity sequential read performance.
Some quick benchmarks with 'cat'-ing a 250MB file after dropping caches:
On an ARM64 phone (using sha256-ce):
Before: 217 MB/s
After: 263 MB/s
(compare to sha256sum of non-verity file: 357 MB/s)
In an x86_64 VM (using sha256-avx2):
Before: 173 MB/s
After: 215 MB/s
(compare to sha256sum of non-verity file: 223 MB/s)
Link: https://lore.kernel.org/r/20200106205533.137005-1-ebiggers@kernel.org
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-01-07 04:55:33 +08:00
|
|
|
valid = verify_page(inode, vi, req, page, 0);
|
2019-07-23 00:26:22 +08:00
|
|
|
|
2020-01-01 01:55:45 +08:00
|
|
|
fsverity_free_hash_request(vi->tree_params.hash_alg, req);
|
2019-07-23 00:26:22 +08:00
|
|
|
|
|
|
|
return valid;
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(fsverity_verify_page);
|
|
|
|
|
|
|
|
#ifdef CONFIG_BLOCK
|
|
|
|
/**
|
|
|
|
* fsverity_verify_bio() - verify a 'read' bio that has just completed
|
2020-05-12 03:21:17 +08:00
|
|
|
* @bio: the bio to verify
|
2019-07-23 00:26:22 +08:00
|
|
|
*
|
|
|
|
* Verify a set of pages that have just been read from a verity file. The pages
|
|
|
|
* must be pagecache pages that are still locked and not yet uptodate. Pages
|
|
|
|
* that fail verification are set to the Error state. Verification is skipped
|
|
|
|
* for pages already in the Error state, e.g. due to fscrypt decryption failure.
|
|
|
|
*
|
|
|
|
* This is a helper function for use by the ->readpages() method of filesystems
|
|
|
|
* that issue bios to read data directly into the page cache. Filesystems that
|
|
|
|
* populate the page cache without issuing bios (e.g. non block-based
|
|
|
|
* filesystems) must instead call fsverity_verify_page() directly on each page.
|
|
|
|
* All filesystems must also call fsverity_verify_page() on holes.
|
|
|
|
*/
|
|
|
|
void fsverity_verify_bio(struct bio *bio)
|
|
|
|
{
|
|
|
|
struct inode *inode = bio_first_page_all(bio)->mapping->host;
|
|
|
|
const struct fsverity_info *vi = inode->i_verity_info;
|
fs-verity: implement readahead of Merkle tree pages
When fs-verity verifies data pages, currently it reads each Merkle tree
page synchronously using read_mapping_page().
Therefore, when the Merkle tree pages aren't already cached, fs-verity
causes an extra 4 KiB I/O request for every 512 KiB of data (assuming
that the Merkle tree uses SHA-256 and 4 KiB blocks). This results in
more I/O requests and performance loss than is strictly necessary.
Therefore, implement readahead of the Merkle tree pages.
For simplicity, we take advantage of the fact that the kernel already
does readahead of the file's *data*, just like it does for any other
file. Due to this, we don't really need a separate readahead state
(struct file_ra_state) just for the Merkle tree, but rather we just need
to piggy-back on the existing data readahead requests.
We also only really need to bother with the first level of the Merkle
tree, since the usual fan-out factor is 128, so normally over 99% of
Merkle tree I/O requests are for the first level.
Therefore, make fsverity_verify_bio() enable readahead of the first
Merkle tree level, for up to 1/4 the number of pages in the bio, when it
sees that the REQ_RAHEAD flag is set on the bio. The readahead size is
then passed down to ->read_merkle_tree_page() for the filesystem to
(optionally) implement if it sees that the requested page is uncached.
While we're at it, also make build_merkle_tree_level() set the Merkle
tree readahead size, since it's easy to do there.
However, for now don't set the readahead size in fsverity_verify_page(),
since currently it's only used to verify holes on ext4 and f2fs, and it
would need parameters added to know how much to read ahead.
This patch significantly improves fs-verity sequential read performance.
Some quick benchmarks with 'cat'-ing a 250MB file after dropping caches:
On an ARM64 phone (using sha256-ce):
Before: 217 MB/s
After: 263 MB/s
(compare to sha256sum of non-verity file: 357 MB/s)
In an x86_64 VM (using sha256-avx2):
Before: 173 MB/s
After: 215 MB/s
(compare to sha256sum of non-verity file: 223 MB/s)
Link: https://lore.kernel.org/r/20200106205533.137005-1-ebiggers@kernel.org
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-01-07 04:55:33 +08:00
|
|
|
const struct merkle_tree_params *params = &vi->tree_params;
|
2019-07-23 00:26:22 +08:00
|
|
|
struct ahash_request *req;
|
|
|
|
struct bio_vec *bv;
|
|
|
|
struct bvec_iter_all iter_all;
|
fs-verity: implement readahead of Merkle tree pages
When fs-verity verifies data pages, currently it reads each Merkle tree
page synchronously using read_mapping_page().
Therefore, when the Merkle tree pages aren't already cached, fs-verity
causes an extra 4 KiB I/O request for every 512 KiB of data (assuming
that the Merkle tree uses SHA-256 and 4 KiB blocks). This results in
more I/O requests and performance loss than is strictly necessary.
Therefore, implement readahead of the Merkle tree pages.
For simplicity, we take advantage of the fact that the kernel already
does readahead of the file's *data*, just like it does for any other
file. Due to this, we don't really need a separate readahead state
(struct file_ra_state) just for the Merkle tree, but rather we just need
to piggy-back on the existing data readahead requests.
We also only really need to bother with the first level of the Merkle
tree, since the usual fan-out factor is 128, so normally over 99% of
Merkle tree I/O requests are for the first level.
Therefore, make fsverity_verify_bio() enable readahead of the first
Merkle tree level, for up to 1/4 the number of pages in the bio, when it
sees that the REQ_RAHEAD flag is set on the bio. The readahead size is
then passed down to ->read_merkle_tree_page() for the filesystem to
(optionally) implement if it sees that the requested page is uncached.
While we're at it, also make build_merkle_tree_level() set the Merkle
tree readahead size, since it's easy to do there.
However, for now don't set the readahead size in fsverity_verify_page(),
since currently it's only used to verify holes on ext4 and f2fs, and it
would need parameters added to know how much to read ahead.
This patch significantly improves fs-verity sequential read performance.
Some quick benchmarks with 'cat'-ing a 250MB file after dropping caches:
On an ARM64 phone (using sha256-ce):
Before: 217 MB/s
After: 263 MB/s
(compare to sha256sum of non-verity file: 357 MB/s)
In an x86_64 VM (using sha256-avx2):
Before: 173 MB/s
After: 215 MB/s
(compare to sha256sum of non-verity file: 223 MB/s)
Link: https://lore.kernel.org/r/20200106205533.137005-1-ebiggers@kernel.org
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-01-07 04:55:33 +08:00
|
|
|
unsigned long max_ra_pages = 0;
|
2019-07-23 00:26:22 +08:00
|
|
|
|
2020-01-01 01:55:45 +08:00
|
|
|
/* This allocation never fails, since it's mempool-backed. */
|
|
|
|
req = fsverity_alloc_hash_request(params->hash_alg, GFP_NOFS);
|
2019-07-23 00:26:22 +08:00
|
|
|
|
fs-verity: implement readahead of Merkle tree pages
When fs-verity verifies data pages, currently it reads each Merkle tree
page synchronously using read_mapping_page().
Therefore, when the Merkle tree pages aren't already cached, fs-verity
causes an extra 4 KiB I/O request for every 512 KiB of data (assuming
that the Merkle tree uses SHA-256 and 4 KiB blocks). This results in
more I/O requests and performance loss than is strictly necessary.
Therefore, implement readahead of the Merkle tree pages.
For simplicity, we take advantage of the fact that the kernel already
does readahead of the file's *data*, just like it does for any other
file. Due to this, we don't really need a separate readahead state
(struct file_ra_state) just for the Merkle tree, but rather we just need
to piggy-back on the existing data readahead requests.
We also only really need to bother with the first level of the Merkle
tree, since the usual fan-out factor is 128, so normally over 99% of
Merkle tree I/O requests are for the first level.
Therefore, make fsverity_verify_bio() enable readahead of the first
Merkle tree level, for up to 1/4 the number of pages in the bio, when it
sees that the REQ_RAHEAD flag is set on the bio. The readahead size is
then passed down to ->read_merkle_tree_page() for the filesystem to
(optionally) implement if it sees that the requested page is uncached.
While we're at it, also make build_merkle_tree_level() set the Merkle
tree readahead size, since it's easy to do there.
However, for now don't set the readahead size in fsverity_verify_page(),
since currently it's only used to verify holes on ext4 and f2fs, and it
would need parameters added to know how much to read ahead.
This patch significantly improves fs-verity sequential read performance.
Some quick benchmarks with 'cat'-ing a 250MB file after dropping caches:
On an ARM64 phone (using sha256-ce):
Before: 217 MB/s
After: 263 MB/s
(compare to sha256sum of non-verity file: 357 MB/s)
In an x86_64 VM (using sha256-avx2):
Before: 173 MB/s
After: 215 MB/s
(compare to sha256sum of non-verity file: 223 MB/s)
Link: https://lore.kernel.org/r/20200106205533.137005-1-ebiggers@kernel.org
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-01-07 04:55:33 +08:00
|
|
|
if (bio->bi_opf & REQ_RAHEAD) {
|
|
|
|
/*
|
|
|
|
* If this bio is for data readahead, then we also do readahead
|
|
|
|
* of the first (largest) level of the Merkle tree. Namely,
|
|
|
|
* when a Merkle tree page is read, we also try to piggy-back on
|
|
|
|
* some additional pages -- up to 1/4 the number of data pages.
|
|
|
|
*
|
|
|
|
* This improves sequential read performance, as it greatly
|
|
|
|
* reduces the number of I/O requests made to the Merkle tree.
|
|
|
|
*/
|
|
|
|
bio_for_each_segment_all(bv, bio, iter_all)
|
|
|
|
max_ra_pages++;
|
|
|
|
max_ra_pages /= 4;
|
|
|
|
}
|
|
|
|
|
2019-07-23 00:26:22 +08:00
|
|
|
bio_for_each_segment_all(bv, bio, iter_all) {
|
|
|
|
struct page *page = bv->bv_page;
|
fs-verity: implement readahead of Merkle tree pages
When fs-verity verifies data pages, currently it reads each Merkle tree
page synchronously using read_mapping_page().
Therefore, when the Merkle tree pages aren't already cached, fs-verity
causes an extra 4 KiB I/O request for every 512 KiB of data (assuming
that the Merkle tree uses SHA-256 and 4 KiB blocks). This results in
more I/O requests and performance loss than is strictly necessary.
Therefore, implement readahead of the Merkle tree pages.
For simplicity, we take advantage of the fact that the kernel already
does readahead of the file's *data*, just like it does for any other
file. Due to this, we don't really need a separate readahead state
(struct file_ra_state) just for the Merkle tree, but rather we just need
to piggy-back on the existing data readahead requests.
We also only really need to bother with the first level of the Merkle
tree, since the usual fan-out factor is 128, so normally over 99% of
Merkle tree I/O requests are for the first level.
Therefore, make fsverity_verify_bio() enable readahead of the first
Merkle tree level, for up to 1/4 the number of pages in the bio, when it
sees that the REQ_RAHEAD flag is set on the bio. The readahead size is
then passed down to ->read_merkle_tree_page() for the filesystem to
(optionally) implement if it sees that the requested page is uncached.
While we're at it, also make build_merkle_tree_level() set the Merkle
tree readahead size, since it's easy to do there.
However, for now don't set the readahead size in fsverity_verify_page(),
since currently it's only used to verify holes on ext4 and f2fs, and it
would need parameters added to know how much to read ahead.
This patch significantly improves fs-verity sequential read performance.
Some quick benchmarks with 'cat'-ing a 250MB file after dropping caches:
On an ARM64 phone (using sha256-ce):
Before: 217 MB/s
After: 263 MB/s
(compare to sha256sum of non-verity file: 357 MB/s)
In an x86_64 VM (using sha256-avx2):
Before: 173 MB/s
After: 215 MB/s
(compare to sha256sum of non-verity file: 223 MB/s)
Link: https://lore.kernel.org/r/20200106205533.137005-1-ebiggers@kernel.org
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-01-07 04:55:33 +08:00
|
|
|
unsigned long level0_index = page->index >> params->log_arity;
|
|
|
|
unsigned long level0_ra_pages =
|
|
|
|
min(max_ra_pages, params->level0_blocks - level0_index);
|
2019-07-23 00:26:22 +08:00
|
|
|
|
fs-verity: implement readahead of Merkle tree pages
When fs-verity verifies data pages, currently it reads each Merkle tree
page synchronously using read_mapping_page().
Therefore, when the Merkle tree pages aren't already cached, fs-verity
causes an extra 4 KiB I/O request for every 512 KiB of data (assuming
that the Merkle tree uses SHA-256 and 4 KiB blocks). This results in
more I/O requests and performance loss than is strictly necessary.
Therefore, implement readahead of the Merkle tree pages.
For simplicity, we take advantage of the fact that the kernel already
does readahead of the file's *data*, just like it does for any other
file. Due to this, we don't really need a separate readahead state
(struct file_ra_state) just for the Merkle tree, but rather we just need
to piggy-back on the existing data readahead requests.
We also only really need to bother with the first level of the Merkle
tree, since the usual fan-out factor is 128, so normally over 99% of
Merkle tree I/O requests are for the first level.
Therefore, make fsverity_verify_bio() enable readahead of the first
Merkle tree level, for up to 1/4 the number of pages in the bio, when it
sees that the REQ_RAHEAD flag is set on the bio. The readahead size is
then passed down to ->read_merkle_tree_page() for the filesystem to
(optionally) implement if it sees that the requested page is uncached.
While we're at it, also make build_merkle_tree_level() set the Merkle
tree readahead size, since it's easy to do there.
However, for now don't set the readahead size in fsverity_verify_page(),
since currently it's only used to verify holes on ext4 and f2fs, and it
would need parameters added to know how much to read ahead.
This patch significantly improves fs-verity sequential read performance.
Some quick benchmarks with 'cat'-ing a 250MB file after dropping caches:
On an ARM64 phone (using sha256-ce):
Before: 217 MB/s
After: 263 MB/s
(compare to sha256sum of non-verity file: 357 MB/s)
In an x86_64 VM (using sha256-avx2):
Before: 173 MB/s
After: 215 MB/s
(compare to sha256sum of non-verity file: 223 MB/s)
Link: https://lore.kernel.org/r/20200106205533.137005-1-ebiggers@kernel.org
Reviewed-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Eric Biggers <ebiggers@google.com>
2020-01-07 04:55:33 +08:00
|
|
|
if (!PageError(page) &&
|
|
|
|
!verify_page(inode, vi, req, page, level0_ra_pages))
|
2019-07-23 00:26:22 +08:00
|
|
|
SetPageError(page);
|
|
|
|
}
|
|
|
|
|
2020-01-01 01:55:45 +08:00
|
|
|
fsverity_free_hash_request(params->hash_alg, req);
|
2019-07-23 00:26:22 +08:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(fsverity_verify_bio);
|
|
|
|
#endif /* CONFIG_BLOCK */
|
|
|
|
|
|
|
|
/**
|
|
|
|
* fsverity_enqueue_verify_work() - enqueue work on the fs-verity workqueue
|
2020-05-12 03:21:17 +08:00
|
|
|
* @work: the work to enqueue
|
2019-07-23 00:26:22 +08:00
|
|
|
*
|
|
|
|
* Enqueue verification work for asynchronous processing.
|
|
|
|
*/
|
|
|
|
void fsverity_enqueue_verify_work(struct work_struct *work)
|
|
|
|
{
|
|
|
|
queue_work(fsverity_read_workqueue, work);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(fsverity_enqueue_verify_work);
|
|
|
|
|
|
|
|
int __init fsverity_init_workqueue(void)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* Use an unbound workqueue to allow bios to be verified in parallel
|
|
|
|
* even when they happen to complete on the same CPU. This sacrifices
|
|
|
|
* locality, but it's worthwhile since hashing is CPU-intensive.
|
|
|
|
*
|
|
|
|
* Also use a high-priority workqueue to prioritize verification work,
|
|
|
|
* which blocks reads from completing, over regular application tasks.
|
|
|
|
*/
|
|
|
|
fsverity_read_workqueue = alloc_workqueue("fsverity_read_queue",
|
|
|
|
WQ_UNBOUND | WQ_HIGHPRI,
|
|
|
|
num_online_cpus());
|
|
|
|
if (!fsverity_read_workqueue)
|
|
|
|
return -ENOMEM;
|
|
|
|
return 0;
|
|
|
|
}
|
2019-07-23 00:26:23 +08:00
|
|
|
|
|
|
|
void __init fsverity_exit_workqueue(void)
|
|
|
|
{
|
|
|
|
destroy_workqueue(fsverity_read_workqueue);
|
|
|
|
fsverity_read_workqueue = NULL;
|
|
|
|
}
|