c/mirror_zfs
1
0
Fork 0
mirror of https://git.proxmox.com/git/mirror_zfs.git synced 2026-05-22 18:40:43 +03:00
Code Issues Packages Projects Releases Wiki Activity

ICP: Improve AES-GCM performance

Browse Source 
Currently SIMD accelerated AES-GCM performance is limited by two
factors:

a. The need to disable preemption and interrupts and save the FPU
state before using it and to do the reverse when done. Due to the
way the code is organized (see (b) below) we have to pay this price
twice for each 16 byte GCM block processed.

b. Most processing is done in C, operating on single GCM blocks.
The use of SIMD instructions is limited to the AES encryption of the
counter block (AES-NI) and the Galois multiplication (PCLMULQDQ).
This leads to the FPU not being fully utilized for crypto
operations.

To solve (a) we do crypto processing in larger chunks while owning
the FPU. An `icp_gcm_avx_chunk_size` module parameter was introduced
to make this chunk size tweakable. It defaults to 32 KiB. This step
alone roughly doubles performance. (b) is tackled by porting and
using the highly optimized openssl AES-GCM assembler routines, which
do all the processing (CTR, AES, GMULT) in a single routine. Both
steps together result in up to 32x reduction of the time spend in
the en/decryption routines, leading up to approximately 12x
throughput increase for large (128 KiB) blocks.

Lastly, this commit changes the default encryption algorithm from
AES-CCM to AES-GCM when setting the `encryption=on` property.

Reviewed-By: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-By: Jason King <jason.king@joyent.com>
Reviewed-By: Tom Caputi <tcaputi@datto.com>
Reviewed-By: Richard Laager <rlaager@wiktel.com>
Signed-off-by: Attila Fülöp <attila@fueloep.org>
Closes #9749
This commit is contained in:
Attila Fülöp
2020-02-10 21:59:50 +01:00
committed by Tony Hutter
parent 590ababea2
commit 76354f945e
20 changed files with 2657 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files
module/icp/include/modes/modes.h
+27 -2
View File
@@ -34,6 +34,16 @@ extern "C" {
#include <sys/crypto/common.h>
#include <sys/crypto/impl.h>
/*
* Does the build chain support all instructions needed for the GCM assembler
* routines. AVX support should imply AES-NI and PCLMULQDQ, but make sure
* anyhow.
*/
#if defined(__x86_64__) && defined(HAVE_AVX) && \
defined(HAVE_AES) && defined(HAVE_PCLMULQDQ) && defined(HAVE_MOVBE)
#define CAN_USE_GCM_ASM
#endif
#define ECB_MODE 0x00000002
#define CBC_MODE 0x00000004
#define CTR_MODE 0x00000008
@@ -189,13 +199,17 @@ typedef struct ccm_ctx {
*
* gcm_H: Subkey.
*
* gcm_Htable: Pre-computed and pre-shifted H, H^2, ... H^6 for the
* Karatsuba Algorithm in host byte order.
*
* gcm_J0: Pre-counter block generated from the IV.
*
* gcm_len_a_len_c: 64-bit representations of the bit lengths of
* AAD and ciphertext.
*
* gcm_kmflag: Current value of kmflag. Used only for allocating
* the plaintext buffer during decryption.
* gcm_kmflag: Current value of kmflag. Used for allocating
* the plaintext buffer during decryption and a
* gcm_avx_chunk_size'd buffer for avx enabled encryption.
*/
typedef struct gcm_ctx {
struct common_ctx gcm_common;
@@ -203,12 +217,23 @@ typedef struct gcm_ctx {
size_t gcm_processed_data_len;
size_t gcm_pt_buf_len;
uint32_t gcm_tmp[4];
/*
* The relative positions of gcm_ghash, gcm_H and pre-computed
* gcm_Htable are hard coded in aesni-gcm-x86_64.S and ghash-x86_64.S,
* so please don't change (or adjust accordingly).
*/
uint64_t gcm_ghash[2];
uint64_t gcm_H[2];
#ifdef CAN_USE_GCM_ASM
uint64_t gcm_Htable[12][2];
#endif
uint64_t gcm_J0[2];
uint64_t gcm_len_a_len_c[2];
uint8_t *gcm_pt_buf;
int gcm_kmflag;
#ifdef CAN_USE_GCM_ASM
boolean_t gcm_use_avx;
#endif
} gcm_ctx_t;
#define gcm_keysched gcm_common.cc_keysched