Skip to contentSkip to conversation

LZF Cryptoapi support.

PATCH Nigel Cunningham 3 participants 5 messages Jul 18, 2005 – Jul 31, 2005 Crypto API

lore
Share & tools
lore
#0 lore raw

Hi again.

Here's another resend, this time adding an lzf cryptoapi module.

LZF is a very fast compressor which typically achieves approximately 50% compression on a suspend image. The original author (Marc Alexander Lehmann) donated it to Suspend2. I have converted it to cryptoapi with a recent switch of Suspend2 to use cryptoapi.

Regards,

Nigel

 Kconfig  |    7 +
 Makefile |    1 
 lzf.c    |  425 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 433 insertions(+)
crypto/Kconfig +7 -0
diff -ruNp 625-crypto-api-work.patch-old/crypto/Kconfig 625-crypto-api-work.patch-new/crypto/Kconfig
@@ -262,6 +262,13 @@ config CRYPTO_DEFLATE
 	  
 	  You will most probably want this if using IPSec.
 
+config CRYPTO_LZF
+	tristate "LZF compression algorithm"
+	depends on CRYPTO
+	help
+	  This is the LZF algorithm. It is especially useful for Suspend2,
+	  because it achieves good compression quickly.
+
 config CRYPTO_MICHAEL_MIC
 	tristate "Michael MIC keyed digest algorithm"
 	depends on CRYPTO
crypto/lzf.c +425 -0

425 lines not on this page · raw message

crypto/Makefile +1 -0
diff -ruNp 625-crypto-api-work.patch-old/crypto/Makefile 625-crypto-api-work.patch-new/crypto/Makefile
@@ -30,5 +30,6 @@ obj-$(CONFIG_CRYPTO_ANUBIS) += anubis.o
 obj-$(CONFIG_CRYPTO_DEFLATE) += deflate.o
 obj-$(CONFIG_CRYPTO_MICHAEL_MIC) += michael_mic.o
 obj-$(CONFIG_CRYPTO_CRC32C) += crc32c.o
+obj-$(CONFIG_CRYPTO_LZF) += lzf.o
 
 obj-$(CONFIG_CRYPTO_TEST) += tcrypt.o
Signature · 5 lines
Evolution.
Enumerate the requirements.
Consider the interdependencies.
Calculate the probabilities.
Be amazed that people believe it happened. 
#1 lore raw

Hi!

Quoted 6 lines

Here's another resend, this time adding an lzf cryptoapi module.

LZF is a very fast compressor which typically achieves approximately 50% compression on a suspend image. The original author (Marc Alexander Lehmann) donated it to Suspend2. I have converted it to cryptoapi with a recent switch of Suspend2 to use cryptoapi.

And it is still that old, ugly code.

crypto/lzf.c line 91

Quoted 11 lines
crypto/lzf.c +11 -0
+/*
+ * sacrifice some compression quality in favour of compression speed.
+ * (roughly 1-2% worse compression for large blocks and
+ * 9-10% for small, redundant, blocks and >>20% better speed in both cases)
+ * In short: enable this for binary data, disable this for text data.
+ */
+#define ULTRA_FAST 1
+
+#define STRICT_ALIGN 0
+#define USE_MEMCPY 1
+#define INIT_HTAB 0

We do not want these options. It also allows you to kill ifdefs down in the code.

Quoted 14 lines
crypto/lzf.c +14 -0
+#define HSIZE (1 << (HLOG))
+
+/*
+ * don't play with this unless you benchmark!
+ * decompression is not dependent on the hash function
+ * the hashing function might seem strange, just believe me
+ * it works ;)
+ */
+#define FRST(p) (((p[0]) << 8) + p[1])
+#define NEXT(v,p) (((v) << 8) + p[2])
+#define IDX(h) ((((h ^ (h << 5)) >> (3*8 - HLOG)) + h*3) & (HSIZE - 1))
+/*
+ * IDX works because it is very similar to a multiplicative hash, e.g.
+ * (h * 57321 >> (3*8 - HLOG))

crypto/lzf.c line 116

Quoted 10 lines
crypto/lzf.c +10 -0
+ * the next one is also quite good, albeit slow ;)
+ * (int)(cos(h & 0xffffff) * 1e6)
+ */
+
+#if 0
+/* original lzv-like hash function */
+# define FRST(p) (p[0] << 5) ^ p[1]
+# define NEXT(v,p) ((v) << 5) ^ p[2]
+# define IDX(h) ((h) & (HSIZE - 1))
+#endif

And we do not want #if 0-ed code.

Pavel

Signature · 1 line
teflon -- maybe it is a trademark, but it should not be.
#2 lore raw

Howdy.

New revision. Anything left to fix up?

Regards,

Nigel

 Kconfig  |    7 +
 Makefile |    1 
 lzf.c    |  377 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 385 insertions(+)
crypto/Kconfig +7 -0
diff -ruNp 260-crypto-api-lzf-support.patch-old/crypto/Kconfig 260-crypto-api-lzf-support.patch-new/crypto/Kconfig
@@ -282,6 +282,13 @@ config CRYPTO_DEFLATE
 	  
 	  You will most probably want this if using IPSec.
 
+config CRYPTO_LZF
+	tristate "LZF compression algorithm"
+	depends on CRYPTO
+	help
+	  This is the LZF algorithm. It is especially useful for Suspend2,
+	  because it achieves good compression quickly.
+
 config CRYPTO_MICHAEL_MIC
 	tristate "Michael MIC keyed digest algorithm"
 	depends on CRYPTO
crypto/lzf.c +377 -0

377 lines not on this page · raw message

crypto/Makefile +1 -0
diff -ruNp 260-crypto-api-lzf-support.patch-old/crypto/Makefile 260-crypto-api-lzf-support.patch-new/crypto/Makefile
@@ -30,5 +30,6 @@ obj-$(CONFIG_CRYPTO_ANUBIS) += anubis.o
 obj-$(CONFIG_CRYPTO_DEFLATE) += deflate.o
 obj-$(CONFIG_CRYPTO_MICHAEL_MIC) += michael_mic.o
 obj-$(CONFIG_CRYPTO_CRC32C) += crc32c.o
+obj-$(CONFIG_CRYPTO_LZF) += lzf.o
 
 obj-$(CONFIG_CRYPTO_TEST) += tcrypt.o
Signature · 4 lines
Evolution.
Enumerate the requirements.
Consider the interdependencies.
Calculate the probabilities.
#3 lore raw

Hi!

Quoted 1 line

New revision. Anything left to fix up?

It certainly looks better now.

Quoted 13 lines
crypto/lzf.c +13 -0
+static int lzf_compress_init(void *context)
+{
+	struct lzf_ctx *ctx = (struct lzf_ctx *)context;
+
+	/* Get LZF ready to go */
+	ctx->hbuf = vmalloc_32((1 << hlog) * sizeof(char *));
+	if (!ctx->hbuf) {
+		printk(KERN_WARNING
+		       "Failed to allocate %d bytes for lzf workspace\n",
+		       (1 << hlog) * sizeof(char *));
+		return -ENOMEM;
+	}
+

crypto/lzf.c line 166

Quoted 10 lines
crypto/lzf.c +10 -0
+	/* Allocate local buffer */
+	ctx->local_buffer = (char *)get_zeroed_page(GFP_ATOMIC);
+
+	if (!ctx->local_buffer) {
+		lzf_compress_exit(ctx);
+		return -ENOMEM;
+	}
+
+	/* Allocate page buffer */
+	ctx->page_buffer = (char *)get_zeroed_page(GFP_ATOMIC);

Why GFP_ATOMIC in last two?

crypto/lzf.c line 179

Quoted 12 lines
crypto/lzf.c +12 -0
+	if (!ctx->page_buffer) {
+		free_page((unsigned long)ctx->local_buffer);
+		lzf_compress_exit(ctx);
+		return -ENOMEM;
+	}
+
+	ctx->first_call = 1;
+
+	return 0;
+}
+
+static int lzf_compress(void *context, const u8 * in_data, unsigned int in_len,

~ extra space; it
is inconsistent all over the file.

Pavel

Signature · 1 line
Boycott Kodak -- for their patent abuse against Java.
maintainer
#4 lore raw
In reply to Nigel Cunningham

crypto/lzf.c line 149

Quoted 7 lines · Nigel Cunningham
crypto/lzf.c +6 -0
+static int lzf_compress_init(void *context)
+{
+	struct lzf_ctx *ctx = (struct lzf_ctx *)context;
+
+	/* Get LZF ready to go */
+	ctx->hbuf = vmalloc_32((1 << hlog) * sizeof(char *));

Any reason why vmalloc can't be used?

crypto/lzf.c line 166

Quoted 5 lines
crypto/lzf.c +4 -0
+	/* Allocate local buffer */
+	ctx->local_buffer = (char *)get_zeroed_page(GFP_ATOMIC);
 
+	/* Allocate page buffer */
+	ctx->page_buffer = (char *)get_zeroed_page(GFP_ATOMIC);

Where are these two buffers used anyway?

crypto/lzf.c line 170

Quoted 3 lines
crypto/lzf.c +3 -0
+	if (!ctx->page_buffer) {
+		free_page((unsigned long)ctx->local_buffer);
+		lzf_compress_exit(ctx);

This is a double-free of local_buffer.

Cheers,

Signature · 4 lines
Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt