coreboot-kgpe-d16/payloads/libpayload/liblz4/lz4_wrapper.c

163 lines
5.1 KiB
C
Raw Normal View History

libpayload: Add LZ4 decompression algorithm This patch adds support for the LZ4 decompression algorithm to libpayload. It's what all the cool kids are using for decompression these days and has many interesting advantages over LZMA (and everything else I know of): blazing fast decompression (20(!) times faster than LZMA, twice as fast as LZO on my Cortex-A72), no memory requirements on decompression, and possibly in-place decompression support. It pays for that with a lower compression ratio (about 50% larger compressed size than LZMA, 10% larger than LZO for an ARM64 Linux kernel binary), but the boot time math still works in its favor for our IO speeds. This patch only adds the raw decompression functions for use by external payloads, we can later try integrating them in CBFS. It copies the decompression code itself unmodified from the upstream LZ4 library at github.com/Cyan4973/lz4 which will hopefully make it easy to update. The frame format parsing is reimplemented since the upstream version looks unnecessarily complex and unreadable for our needs. BRANCH=smaug BUG=chrome-os-partner:32184 TEST=With other patches, booted ARM64 kernel that got compressed from 15M to 5.1M and decompresses in 44ms. Change-Id: I65bdc4b2b19bd51c7b7e17a4e4b79da301a2a014 Signed-off-by: Patrick Georgi <pgeorgi@chromium.org> Original-Commit-Id: f8a1fc996d5b0234d07f567fa8163d0f802d5144 Original-Change-Id: I15c0620da05561ade2552b15ffdf6bb3afd7eb26 Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/282743 Original-Reviewed-by: Stefan Reinauer <reinauer@google.com> Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: http://review.coreboot.org/10845 Tested-by: build bot (Jenkins) Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
2015-06-30 19:30:30 +02:00
/*
* Copyright 2015 Google Inc.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. The name of the author may not be used to endorse or promote products
* derived from this software without specific prior written permission.
*
* Alternatively, this software may be distributed under the terms of the
* GNU General Public License ("GPL") version 2 as published by the Free
* Software Foundation.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
#include <assert.h>
#include <endian.h>
#include <libpayload.h>
#include <lz4.h>
/* LZ4 comes with its own supposedly portable memory access functions, but they
* seem to be very inefficient in practice (at least on ARM64). Since libpayload
* knows about endinaness and allows some basic assumptions (such as unaligned
* access support), we can easily write the ones we need ourselves. */
static u16 LZ4_readLE16(const void *src) { return le16toh(*(u16 *)src); }
static void LZ4_copy4(void *dst, const void *src) { *(u32 *)dst = *(u32 *)src; }
static void LZ4_copy8(void *dst, const void *src) { *(u64 *)dst = *(u64 *)src; }
typedef uint8_t BYTE;
typedef uint16_t U16;
typedef uint32_t U32;
typedef int32_t S32;
typedef uint64_t U64;
#define FORCE_INLINE static inline __attribute__((always_inline))
#define likely(expr) __builtin_expect((expr) != 0, 1)
#define unlikely(expr) __builtin_expect((expr) != 0, 0)
/* Unaltered (except removing unrelated code) from github.com/Cyan4973/lz4. */
#include "lz4.c" /* #include for inlining, do not link! */
#define LZ4F_MAGICNUMBER 0x184D2204
struct lz4_frame_header {
u32 magic;
union {
u8 flags;
struct {
u8 reserved0 : 2;
u8 has_content_checksum : 1;
u8 has_content_size : 1;
u8 has_block_checksum : 1;
u8 independent_blocks : 1;
u8 version : 2;
};
};
union {
u8 block_descriptor;
struct {
u8 reserved1 : 4;
u8 max_block_size : 3;
u8 reserved2 : 1;
};
};
/* + u64 content_size iff has_content_size is set */
/* + u8 header_checksum */
} __attribute__((packed));
struct lz4_block_header {
union {
u32 raw;
struct {
u32 size : 31;
u32 not_compressed : 1;
};
};
/* + size bytes of data */
/* + u32 block_checksum iff has_block_checksum is set */
} __attribute__((packed));
size_t ulz4fn(const void *src, size_t srcn, void *dst, size_t dstn)
{
const void *in = src;
void *out = dst;
int has_block_checksum;
{ /* With in-place decompression the header may become invalid later. */
const struct lz4_frame_header *h = in;
if (srcn < sizeof(*h) + sizeof(u64) + sizeof(u8))
return 0; /* input overrun */
/* We assume there's always only a single, standard frame. */
if (le32toh(h->magic) != LZ4F_MAGICNUMBER || h->version != 1)
return 0; /* unknown format */
if (h->reserved0 || h->reserved1 || h->reserved2)
return 0; /* reserved must be zero */
if (!h->independent_blocks)
return 0; /* we don't support block dependency */
has_block_checksum = h->has_block_checksum;
in += sizeof(*h);
if (h->has_content_size)
in += sizeof(u64);
in += sizeof(u8);
}
while (1) {
struct lz4_block_header b = { .raw = le32toh(*(u32 *)in) };
in += sizeof(struct lz4_block_header);
if (in - src + b.size > srcn)
return 0; /* input overrun */
if (!b.size)
return out - dst; /* decompression successful */
if (b.not_compressed) {
libpayload: lz4: Add output overrun check to incompressible case The LZ4 decompressor currently doesn't check for output overruns before writing data in the case where a block had been incompressible (and included verbatim in the compression stream). This is extremely unlikely with the default 4MB blocks, but still a nice thing to fix. We'll still output as much data as we can before returning an error to support partial decompression use cases. This matches the behavior already in place for normal, LZ4-compressed blocks where the decompression function is already (supposed to be) doing complete bounds checking (although it is not guaranteed to output all valid bytes before aborting on an output overrun, and you should try to provide a few dozen bytes of extra buffer space beyond the parts you're interested in on partial decompression). BRANCH=None BUG=chrome-os-partner:32184 TEST=None Change-Id: I5e40c8cec8947ec0ec8f6d8c8fa2574cfb4dc958 Signed-off-by: Patrick Georgi <pgeorgi@chromium.org> Original-Commit-Id: 636985334c9b3b93a12d4066d2829f1f999c9315 Original-Change-Id: Iecf44650aade60b9fa1b13e57da752fb482a3f3f Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/286240 Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: http://review.coreboot.org/11016 Tested-by: build bot (Jenkins) Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org> Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
2015-07-16 22:59:57 +02:00
size_t size = MIN((u32)b.size, dst + dstn - out);
memcpy(out, in, size);
if (size < b.size)
return 0; /* output overrun */
else
out += size;
libpayload: Add LZ4 decompression algorithm This patch adds support for the LZ4 decompression algorithm to libpayload. It's what all the cool kids are using for decompression these days and has many interesting advantages over LZMA (and everything else I know of): blazing fast decompression (20(!) times faster than LZMA, twice as fast as LZO on my Cortex-A72), no memory requirements on decompression, and possibly in-place decompression support. It pays for that with a lower compression ratio (about 50% larger compressed size than LZMA, 10% larger than LZO for an ARM64 Linux kernel binary), but the boot time math still works in its favor for our IO speeds. This patch only adds the raw decompression functions for use by external payloads, we can later try integrating them in CBFS. It copies the decompression code itself unmodified from the upstream LZ4 library at github.com/Cyan4973/lz4 which will hopefully make it easy to update. The frame format parsing is reimplemented since the upstream version looks unnecessarily complex and unreadable for our needs. BRANCH=smaug BUG=chrome-os-partner:32184 TEST=With other patches, booted ARM64 kernel that got compressed from 15M to 5.1M and decompresses in 44ms. Change-Id: I65bdc4b2b19bd51c7b7e17a4e4b79da301a2a014 Signed-off-by: Patrick Georgi <pgeorgi@chromium.org> Original-Commit-Id: f8a1fc996d5b0234d07f567fa8163d0f802d5144 Original-Change-Id: I15c0620da05561ade2552b15ffdf6bb3afd7eb26 Original-Signed-off-by: Julius Werner <jwerner@chromium.org> Original-Reviewed-on: https://chromium-review.googlesource.com/282743 Original-Reviewed-by: Stefan Reinauer <reinauer@google.com> Original-Reviewed-by: Aaron Durbin <adurbin@chromium.org> Reviewed-on: http://review.coreboot.org/10845 Tested-by: build bot (Jenkins) Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
2015-06-30 19:30:30 +02:00
} else {
/* constant folding essential, do not touch params! */
int ret = LZ4_decompress_generic(in, out, b.size,
dst + dstn - out, endOnInputSize,
full, 0, noDict, out, NULL, 0);
if (ret < 0)
return 0; /* decompression error */
else
out += ret;
}
in += b.size;
if (has_block_checksum)
in += sizeof(u32);
}
}
size_t ulz4f(const void *src, void *dst)
{
/* LZ4 uses signed size parameters, so can't just use ((u32)-1) here. */
return ulz4fn(src, 1*GiB, dst, 1*GiB);
}