FROMLIST: fsverity: improve performance by using multibuffer hashing

When supported by the hash algorithm, use crypto_shash_finup_mb() to
interleave the hashing of pairs of data blocks.  On some CPUs this
nearly doubles hashing performance.  The increase in overall throughput
of cold-cache fsverity reads that I'm seeing on arm64 and x86_64 is
roughly 35% (though this metric is hard to measure as it jumps around a
lot).

For now this is only done on the verification path, and only for data
blocks, not Merkle tree blocks.  We could use finup_mb on Merkle tree
blocks too, but that is less important as there aren't as many Merkle
tree blocks as data blocks, and that would require some additional code
restructuring.  We could also use finup_mb to accelerate building the
Merkle tree, but verification performance is more important.

Reviewed-by: Sami Tolvanen <samitolvanen@google.com>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Eric Biggers <ebiggers@google.com>

Bug: 330611177
Link: https://lore.kernel.org/r/20240621165922.77672-7-ebiggers@kernel.org
Change-Id: Ib5de1fba462825004f46ed0fb344c9a89603405b
Signed-off-by: Eric Biggers <ebiggers@google.com>
This commit is contained in:
Eric Biggers 2024-06-21 10:37:51 -07:00 committed by William McVicker
parent 08600b5d0c
commit abed1a5d36
3 changed files with 154 additions and 32 deletions

View File

@ -29,6 +29,11 @@ struct fsverity_hash_alg {
* FS_VERITY_HASH_ALG_*, which uses a different numbering scheme.
*/
enum hash_algo algo_id;
/*
* The maximum supported interleaving factor for multibuffer hashing, or
* 1 if the algorithm doesn't support multibuffer hashing
*/
int mb_max_msgs;
};
/* Merkle tree parameters: hash algorithm, initial hash state, and topology */
@ -142,6 +147,8 @@ static inline void fsverity_init_signature(void)
/* verify.c */
#define FS_VERITY_MAX_PENDING_DATA_BLOCKS 2
void __init fsverity_init_workqueue(void);
#endif /* _FSVERITY_PRIVATE_H */

View File

@ -84,8 +84,12 @@ const struct fsverity_hash_alg *fsverity_get_hash_alg(const struct inode *inode,
if (WARN_ON_ONCE(alg->block_size != crypto_shash_blocksize(tfm)))
goto err_free_tfm;
pr_info("%s using implementation \"%s\"\n",
alg->name, crypto_shash_driver_name(tfm));
alg->mb_max_msgs = min(crypto_shash_mb_max_msgs(tfm),
FS_VERITY_MAX_PENDING_DATA_BLOCKS);
pr_info("%s using implementation \"%s\"%s\n",
alg->name, crypto_shash_driver_name(tfm),
alg->mb_max_msgs > 1 ? " (multibuffer)" : "");
/* pairs with smp_load_acquire() above */
smp_store_release(&alg->tfm, tfm);

View File

@ -10,6 +10,27 @@
#include <crypto/hash.h>
#include <linux/bio.h>
struct fsverity_pending_block {
const void *data;
u64 pos;
u8 real_hash[FS_VERITY_MAX_DIGEST_SIZE];
};
struct fsverity_verification_context {
struct inode *inode;
struct fsverity_info *vi;
unsigned long max_ra_pages;
/*
* This is the queue of data blocks that are pending verification. We
* allow multiple blocks to be queued up in order to support multibuffer
* hashing, i.e. interleaving the hashing of multiple messages. On many
* CPUs this improves performance significantly.
*/
int num_pending;
struct fsverity_pending_block pending_blocks[FS_VERITY_MAX_PENDING_DATA_BLOCKS];
};
static struct workqueue_struct *fsverity_read_workqueue;
/*
@ -79,7 +100,7 @@ static bool is_hash_block_verified(struct fsverity_info *vi, struct page *hpage,
}
/*
* Verify a single data block against the file's Merkle tree.
* Verify the hash of a single data block 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 blocks. Therefore we need
@ -90,8 +111,10 @@ static bool is_hash_block_verified(struct fsverity_info *vi, struct page *hpage,
*/
static bool
verify_data_block(struct inode *inode, struct fsverity_info *vi,
const void *data, u64 data_pos, unsigned long max_ra_pages)
const struct fsverity_pending_block *dblock,
unsigned long max_ra_pages)
{
const u64 data_pos = dblock->pos;
const struct merkle_tree_params *params = &vi->tree_params;
const unsigned int hsize = params->digest_size;
int level;
@ -115,8 +138,12 @@ verify_data_block(struct inode *inode, struct fsverity_info *vi,
*/
u64 hidx = data_pos >> params->log_blocksize;
/* Up to 1 + FS_VERITY_MAX_LEVELS pages may be mapped at once */
BUILD_BUG_ON(1 + FS_VERITY_MAX_LEVELS > KM_MAX_IDX);
/*
* Up to FS_VERITY_MAX_PENDING_DATA_BLOCKS + FS_VERITY_MAX_LEVELS pages
* may be mapped at once.
*/
BUILD_BUG_ON(FS_VERITY_MAX_PENDING_DATA_BLOCKS +
FS_VERITY_MAX_LEVELS > KM_MAX_IDX);
if (unlikely(data_pos >= inode->i_size)) {
/*
@ -127,7 +154,7 @@ verify_data_block(struct inode *inode, struct fsverity_info *vi,
* any part past EOF should be all zeroes. Therefore, we need
* to verify that any data blocks fully past EOF are all zeroes.
*/
if (memchr_inv(data, 0, params->block_size)) {
if (memchr_inv(dblock->data, 0, params->block_size)) {
fsverity_err(inode,
"FILE CORRUPTED! Data past EOF is not zeroed");
return false;
@ -221,10 +248,8 @@ descend:
put_page(hpage);
}
/* Finally, verify the data block. */
if (fsverity_hash_block(params, inode, data, real_hash) != 0)
goto error;
if (memcmp(want_hash, real_hash, hsize) != 0)
/* Finally, verify the hash of the data block. */
if (memcmp(want_hash, dblock->real_hash, hsize) != 0)
goto corrupted;
return true;
@ -233,7 +258,8 @@ corrupted:
"FILE CORRUPTED! pos=%llu, level=%d, want_hash=%s:%*phN, real_hash=%s:%*phN",
data_pos, level - 1,
params->hash_alg->name, hsize, want_hash,
params->hash_alg->name, hsize, real_hash);
params->hash_alg->name, hsize,
level == 0 ? dblock->real_hash : real_hash);
error:
for (; level > 0; level--) {
kunmap_local(hblocks[level - 1].addr);
@ -242,13 +268,83 @@ error:
return false;
}
static bool
verify_data_blocks(struct folio *data_folio, size_t len, size_t offset,
unsigned long max_ra_pages)
static void
fsverity_init_verification_context(struct fsverity_verification_context *ctx,
struct inode *inode,
unsigned long max_ra_pages)
{
struct inode *inode = data_folio->mapping->host;
struct fsverity_info *vi = inode->i_verity_info;
const unsigned int block_size = vi->tree_params.block_size;
ctx->inode = inode;
ctx->vi = inode->i_verity_info;
ctx->max_ra_pages = max_ra_pages;
ctx->num_pending = 0;
}
static void
fsverity_clear_pending_blocks(struct fsverity_verification_context *ctx)
{
int i;
for (i = ctx->num_pending - 1; i >= 0; i--) {
kunmap_local(ctx->pending_blocks[i].data);
ctx->pending_blocks[i].data = NULL;
}
ctx->num_pending = 0;
}
static bool
fsverity_verify_pending_blocks(struct fsverity_verification_context *ctx)
{
struct inode *inode = ctx->inode;
struct fsverity_info *vi = ctx->vi;
const struct merkle_tree_params *params = &vi->tree_params;
SHASH_DESC_ON_STACK(desc, params->hash_alg->tfm);
const u8 *data[FS_VERITY_MAX_PENDING_DATA_BLOCKS];
u8 *real_hashes[FS_VERITY_MAX_PENDING_DATA_BLOCKS];
int i;
int err;
if (ctx->num_pending == 0)
return true;
for (i = 0; i < ctx->num_pending; i++) {
data[i] = ctx->pending_blocks[i].data;
real_hashes[i] = ctx->pending_blocks[i].real_hash;
}
desc->tfm = params->hash_alg->tfm;
if (params->hashstate)
err = crypto_shash_import(desc, params->hashstate);
else
err = crypto_shash_init(desc);
if (err) {
fsverity_err(inode, "Error %d importing hash state", err);
return false;
}
err = crypto_shash_finup_mb(desc, data, params->block_size, real_hashes,
ctx->num_pending);
if (err) {
fsverity_err(inode, "Error %d computing block hashes", err);
return false;
}
for (i = 0; i < ctx->num_pending; i++) {
if (!verify_data_block(inode, vi, &ctx->pending_blocks[i],
ctx->max_ra_pages))
return false;
}
fsverity_clear_pending_blocks(ctx);
return true;
}
static bool
fsverity_add_data_blocks(struct fsverity_verification_context *ctx,
struct folio *data_folio, size_t len, size_t offset)
{
struct fsverity_info *vi = ctx->vi;
const struct merkle_tree_params *params = &vi->tree_params;
const unsigned int block_size = params->block_size;
const int mb_max_msgs = params->hash_alg->mb_max_msgs;
u64 pos = (u64)data_folio->index << PAGE_SHIFT;
if (WARN_ON_ONCE(len <= 0 || !IS_ALIGNED(len | offset, block_size)))
@ -257,14 +353,11 @@ verify_data_blocks(struct folio *data_folio, size_t len, size_t offset,
folio_test_uptodate(data_folio)))
return false;
do {
void *data;
bool valid;
data = kmap_local_folio(data_folio, offset);
valid = verify_data_block(inode, vi, data, pos + offset,
max_ra_pages);
kunmap_local(data);
if (!valid)
ctx->pending_blocks[ctx->num_pending].data =
kmap_local_folio(data_folio, offset);
ctx->pending_blocks[ctx->num_pending].pos = pos + offset;
if (++ctx->num_pending == mb_max_msgs &&
!fsverity_verify_pending_blocks(ctx))
return false;
offset += block_size;
len -= block_size;
@ -286,7 +379,15 @@ verify_data_blocks(struct folio *data_folio, size_t len, size_t offset,
*/
bool fsverity_verify_blocks(struct folio *folio, size_t len, size_t offset)
{
return verify_data_blocks(folio, len, offset, 0);
struct fsverity_verification_context ctx;
fsverity_init_verification_context(&ctx, folio->mapping->host, 0);
if (fsverity_add_data_blocks(&ctx, folio, len, offset) &&
fsverity_verify_pending_blocks(&ctx))
return true;
fsverity_clear_pending_blocks(&ctx);
return false;
}
EXPORT_SYMBOL_GPL(fsverity_verify_blocks);
@ -307,6 +408,8 @@ EXPORT_SYMBOL_GPL(fsverity_verify_blocks);
*/
void fsverity_verify_bio(struct bio *bio)
{
struct inode *inode = bio_first_folio_all(bio)->mapping->host;
struct fsverity_verification_context ctx;
struct folio_iter fi;
unsigned long max_ra_pages = 0;
@ -323,13 +426,21 @@ void fsverity_verify_bio(struct bio *bio)
max_ra_pages = bio->bi_iter.bi_size >> (PAGE_SHIFT + 2);
}
fsverity_init_verification_context(&ctx, inode, max_ra_pages);
bio_for_each_folio_all(fi, bio) {
if (!verify_data_blocks(fi.folio, fi.length, fi.offset,
max_ra_pages)) {
bio->bi_status = BLK_STS_IOERR;
break;
}
if (!fsverity_add_data_blocks(&ctx, fi.folio, fi.length,
fi.offset))
goto ioerr;
}
if (!fsverity_verify_pending_blocks(&ctx))
goto ioerr;
return;
ioerr:
fsverity_clear_pending_blocks(&ctx);
bio->bi_status = BLK_STS_IOERR;
}
EXPORT_SYMBOL_GPL(fsverity_verify_bio);
#endif /* CONFIG_BLOCK */