erebion/coreboot
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

commonlib/bsd: Add zstd support

Browse source 
This adds the option to compress ramstage and payloads with zstd.

zstd compressed ramstages are typically +5% bigger than lzma compressed
ramstages. The decompressor .text section grows by 20KiB and the
decompressor needs 16KiB more heap than the lzma decompressor.

To use less heap inside the zstd decompressor the build time define
ZSTD_DECODER_INTERNAL_BUFFER is used.

Quote:
 The build macro `ZSTD_DECODER_INTERNAL_BUFFER` can be set to control
 the amount of extra memory used during decompression to store
 literals. This defaults to 64kB.  Reducing this value reduces the
 memory footprint of `ZSTD_DCtx` decompression contexts, but might
 also result in a small decompression speed cost

TEST=Booted on Lenovo X220 with zstd ramstage showed no disadvantage
     over a bigger internal buffer used.
TEST=Booted on Lenovo X220. The zstd decompressor is twice as fast
     as the lzma decompressor.
     cbmem -t shows:
   - finished ZSTD decompress (ignore for x86)         79,444 (24,494)
   - finished LZMA decompress (ignore for x86)         94,971 (45,545)

TEST=Booted on QEMU Q35, QEMU aarch64 virt, QEMU riscv RV64 with
     zstd compressed ramstage.

Change-Id: Ic1b1f53327c598d07bd83d4391e8012d41696a16
Signed-off-by: Arthur Heymans <arthur@aheymans.xyz>
Signed-off-by: Patrick Rudolph <patrick.rudolph@9elements.com>
Reviewed-on: https://review.coreboot.org/c/coreboot/+/69893
Reviewed-by: Julius Werner <jwerner@chromium.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
This commit is contained in:
Arthur Heymans 2022-11-22 16:33:29 +01:00 • committed by Matt DeVillier
commit 2d99da12a9
13 changed files with 147 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

2
src/commonlib/Makefile.mk
View file

@ -1,6 +1,7 @@
## SPDX-License-Identifier: GPL-2.0-only
subdirs-y += storage
subdirs-y += bsd/zstd
bootblock-y += mem_pool.c
verstage-y += mem_pool.c
@ -58,6 +59,7 @@ ramstage-y += bsd/lz4_wrapper.c
postcar-y += bsd/lz4_wrapper.c
all-y += list.c
all-y += bsd/zstd_wrapper.c
ramstage-y += sort.c

2
src/commonlib/bsd/include/commonlib/bsd/compression.h
View file

@ -19,6 +19,6 @@ size_t ulz4fn(const void *src, size_t srcn, void *dst, size_t dstn);
size_t ulz4f(const void *src, void *dst);
/* Decompresses ZSTD image */
size_t uzstdfn(const void *src, size_t srcn, void *dst, size_t dstn);
size_t uzstdn(const void *src, size_t srcn, void *dst, size_t dstn);
#endif /* _COMMONLIB_COMPRESSION_H_ */

43
src/commonlib/bsd/zstd/Makefile.mk Normal file
View file

@ -0,0 +1,43 @@
/* SPDX-License-Identifier: BSD-3-Clause OR GPL-2.0-only */
zstd_support=n
ifeq ($(CONFIG_COMPRESS_RAMSTAGE_ZSTD),y)
zstd_support=y
endif
ifeq ($(CONFIG_COMPRESSED_PAYLOAD_ZSTD),y)
zstd_support=y
endif
ifeq ($(zstd_support),y)
all-y += decompress/huf_decompress.c
all-y += decompress/zstd_ddict.c
all-y += decompress/zstd_decompress_block.c
all-y += decompress/zstd_decompress.c
all-y += common/entropy_common.c
all-y += common/fse_decompress.c
CPPFLAGS_common += -DZSTD_DISABLE_ASM=1
# Measured on Intel Sandy Bridge i5-2540M at 800Mhz:
# Setting code size reduction time loss
# -DHUF_FORCE_DECOMPRESS_X1=1 6.4 KiB 6%
# -DHUF_FORCE_DECOMPRESS_X2=1 5.4 KiB 7%
# -DZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT=1 8.1 KiB 19%
# -DZSTD_FORCE_DECOMPRESS_SEQUENCES_LONG=1 7.8 KiB 15%
CPPFLAGS_common += -DHUF_FORCE_DECOMPRESS_X1=1
CPPFLAGS_common += -DZSTD_FORCE_DECOMPRESS_SEQUENCES_SHORT=1
CPPFLAGS_common += -DZSTD_STRIP_ERROR_STRINGS=1
# Huffman fast decode needs 64bit and LE
CPPFLAGS_x86_32 += -DHUF_DISABLE_FAST_DECODE=1
# Measured on Intel Sandy Bridge i5-2540M at 800Mhz (x86_64 only):
# Setting code size reduction time loss
# -DHUF_DISABLE_FAST_DECODE=1 2.3 KiB 5%
CPPFLAGS_common += -DDYNAMIC_BMI2=0 -DSTATIC_BMI2=0 -DZSTD_DECODER_INTERNAL_BUFFER=2048
endif
CPPFLAGS_common += -I$(src)/commonlib/bsd/zstd
CPPFLAGS_common += -I$(src)/commonlib/bsd/zstd/common
CPPFLAGS_common += -I$(src)/commonlib/bsd/zstd/decompress

2
src/commonlib/bsd/zstd/decompress/zstd_decompress.c
View file

@ -667,7 +667,7 @@ size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
{ const size_t res = ZSTD_decompressFrame(dctx, dst, dstCapacity,
&src, &srcSize);
RETURN_ERROR_IF(
(ZSTD_getErrorCode(res) == ZSTD_error_prefix_unknown)
(ERR_getErrorCode(res) == ZSTD_error_prefix_unknown)
&& (moreThan1Frame==1),
srcSize_wrong,
"At least one frame successfully completed, "

18
src/commonlib/bsd/zstd_wrapper.c Normal file
View file

@ -0,0 +1,18 @@
/* SPDX-License-Identifier: BSD-3-Clause OR GPL-2.0-only */
#include <assert.h>
#include <commonlib/bsd/compression.h>
#include <stdint.h>
#include <stddef.h>
#define ZSTD_STATIC_LINKING_ONLY 1
#include <zstd.h>
#undef ZSTD_STATIC_LINKING_ONLY
#include <zstd_decompress_internal.h>
size_t uzstdn(const void *src, size_t srcn, void *dst, size_t dstn)
{
static ZSTD_DCtx dctx __aligned(8);
if (!ZSTD_initStaticDCtx(&dctx, sizeof(dctx)))
return 0;
return ZSTD_decompressDCtx(&dctx, dst, dstn, src, srcn);
}

4
src/commonlib/include/commonlib/timestamp_serialized.h
View file

@ -37,6 +37,8 @@ enum timestamp_id {
TS_ULZMA_END = 16,
TS_ULZ4F_START = 17,
TS_ULZ4F_END = 18,
TS_UZSTDF_START = 19,
TS_UZSTDF_END = 20,
TS_DEVICE_INIT_CHIPS = 30,
TS_DEVICE_ENUMERATE = 31,
TS_DEVICE_CONFIGURE = 40,
@ -236,6 +238,8 @@ static const struct timestamp_id_to_name {
TS_NAME_DEF(TS_ULZMA_END, 0, "finished LZMA decompress (ignore for x86)"),
TS_NAME_DEF(TS_ULZ4F_START, TS_ULZ4F_END, "starting LZ4 decompress (ignore for x86)"),
TS_NAME_DEF(TS_ULZ4F_END, 0, "finished LZ4 decompress (ignore for x86)"),
TS_NAME_DEF(TS_UZSTDF_START, TS_UZSTDF_END, "starting ZSTD decompress (ignore for x86)"),
TS_NAME_DEF(TS_UZSTDF_END, 0, "finished ZSTD decompress (ignore for x86)"),
TS_NAME_DEF(TS_DEVICE_INIT_CHIPS, TS_DEVICE_ENUMERATE, "early chipset initialization"),
TS_NAME_DEF(TS_DEVICE_ENUMERATE, TS_DEVICE_CONFIGURE, "device enumeration"),
TS_NAME_DEF(TS_DEVICE_CONFIGURE, TS_DEVICE_ENABLE, "device configuration"),