From 12a2b8ed63f5afda49af3114ebebbc84052f8a6b Mon Sep 17 00:00:00 2001
From: Moinak Ghosh <moinakg@gmail.com>
Date: Fri, 30 Aug 2013 19:51:43 +0530
Subject: [PATCH] Additional error checks in RLE encoding for bsdiff extra
 data. Add a buffer overflow check in RLE encoder. Avoid calling RLE encoding
 if extra data length is zero. Make 2KB block size default for non-global
 deduplication. Update test cases for new 2KB block size support.

---
 README.md            |  7 ++++---
 bsdiff/bsdiff.c      | 16 +++++++++++++---
 bsdiff/bspatch.c     | 16 +++++++++++-----
 bsdiff/rle_encoder.c |  1 +
 pcompress.c          |  9 ++++++++-
 test/t4.tst          |  2 +-
 test/t9.tst          |  2 +-
 7 files changed, 39 insertions(+), 14 deletions(-)

diff --git a/README.md b/README.md
index 74eceae..c3c14e3 100644
--- a/README.md
+++ b/README.md
@@ -154,9 +154,10 @@ NOTE: The option "libbsc" uses  Ilya Grebnov's block sorting compression library
                   gives lower dedupe ratio than content-aware dedupe (-D) and does not
                   support delta compression.
 
-       '-B' <1..5>
-            -     Specify an average Dedupe block size. 1 - 4K, 2 - 8K ... 5 - 64K.
-                  Default deduplication block size is 4KB.
+       '-B' <0..5>
+            -     Specify an average Dedupe block size. 0 - 2K, 1 - 4K, 2 - 8K ... 5 - 64K.
+                  Default deduplication block size is 4KB for Global Deduplication and 2KB
+                  otherwise.
        '-B' 0
             -     This uses blocks as small as 2KB for deduplication. This option can be
                   used for datasets of a few GBs to a few hundred TBs in size depending on
diff --git a/bsdiff/bsdiff.c b/bsdiff/bsdiff.c
index f5b9b39..4371de1 100644
--- a/bsdiff/bsdiff.c
+++ b/bsdiff/bsdiff.c
@@ -444,9 +444,19 @@ bsdiff(u_char *oldbuf, bsize_t oldsize, u_char *newbuf, bsize_t newsize,
 	/* Write extra data */
 	len = newsize - rv;
 	ulen = len;
-	if (zero_rle_encode(eb, eblen, BUFPTR(&pf), &ulen) == -1) {
-		rv = 0;
-		goto out;
+	if (eblen > 0) {
+		if (zero_rle_encode(eb, eblen, BUFPTR(&pf), &ulen) == -1) {
+			rv = 0;
+			goto out;
+		}
+		if (ulen >= eblen) {
+			if (eblen > len) {
+				rv = 0;
+				goto out;
+			}
+			memcpy(BUFPTR(&pf), eb, eblen);
+			ulen = eblen;
+		}
 	}
 	/* Output size of extra data */
 	len = ulen;
diff --git a/bsdiff/bspatch.c b/bsdiff/bspatch.c
index 474fe0b..84fd417 100644
--- a/bsdiff/bspatch.c
+++ b/bsdiff/bspatch.c
@@ -184,12 +184,18 @@ bspatch(u_char *pbuf, u_char *oldbuf, bsize_t oldsize, u_char *newbuf, bsize_t *
 	datalen = len;
 
 	len = extralen;
-	if (zero_rle_decode(pbuf + hdrsz + lzctrllen + lzdatalen, lzextralen, extradata, &len) == -1 ||
-	    len != extralen) {
-		fprintf(stderr, "bspatch: Failed to decompress extra data.\n");
-		rv = 0;
-		goto out;
+	if (len > 0) {
+		if (extralen == lzextralen) {
+			memcpy(extradata, pbuf + hdrsz + lzctrllen + lzdatalen, lzextralen);
+
+		} else if (zero_rle_decode(pbuf + hdrsz + lzctrllen + lzdatalen, lzextralen, extradata, &len) == -1 ||
+		    len != extralen) {
+			fprintf(stderr, "bspatch: Failed to decompress extra data.\n");
+			rv = 0;
+			goto out;
+		}
 	}
+
 	extralen = len;
 	BUFOPEN(&cpf, ctrldata, ctrllen);
 	BUFOPEN(&dpf, diffdata, datalen);
diff --git a/bsdiff/rle_encoder.c b/bsdiff/rle_encoder.c
index 02200b5..c5077dc 100644
--- a/bsdiff/rle_encoder.c
+++ b/bsdiff/rle_encoder.c
@@ -61,6 +61,7 @@ zero_rle_encode(const void *ibuf, const unsigned int ilen,
 			count |= ZERO_MASK;
 			*((unsigned short *)(ob + pos2)) = htons(count);
 			pos2 += 2;
+			if (pos2 > *olen) break;
 		} else {
 			unsigned int pos3, pos4, state;
 			pos3 = pos2;
diff --git a/pcompress.c b/pcompress.c
index e5b8bc2..ad4175b 100644
--- a/pcompress.c
+++ b/pcompress.c
@@ -2595,7 +2595,7 @@ init_pc_context(pc_ctx_t *pctx, int argc, char *argv[])
 	pctx->level = -1;
 	err = 0;
 	pctx->keylen = DEFAULT_KEYLEN;
-	pctx->chunksize = DEFAULT_CHUNKSIZE;
+	pctx->chunksize = -1;
 	pos = argv[0] + strlen(argv[0]);
 	while (*pos != '/' && pos > argv[0]) pos--;
 	if (*pos == '/') pos++;
@@ -2760,6 +2760,13 @@ init_pc_context(pc_ctx_t *pctx, int argc, char *argv[])
 			pctx->level = 6;
 		}
 	}
+
+	if (pctx->chunksize == -1) {
+		if (!pctx->enable_rabin_global)
+			pctx->chunksize = 0;
+		else
+			pctx->chunksize = DEFAULT_CHUNKSIZE;
+	}
 	/*
 	 * Remaining mandatory arguments are the filenames.
 	 */
diff --git a/test/t4.tst b/test/t4.tst
index a53adb3..a3ed244 100644
--- a/test/t4.tst
+++ b/test/t4.tst
@@ -10,7 +10,7 @@ do
 	for tf in `cat files.lst`
 	do
 		rm -f ${tf}.*
-		for feat in "-D" "-D -B3 -L" "-D -B4 -E" "-D -B2 -EE" "-D -B5 -EE -L" "-D -B2 -r" "-P" "-D -P" "-D -L -P" \
+		for feat in "-D" "-D -B3 -L" "-D -B4 -E" "-D -B0 -EE" "-D -B5 -EE -L" "-D -B2 -r" "-P" "-D -P" "-D -L -P" \
 				"-G -D" "-G -F" "-G -L -P" "-G -B2"
 		do
 			for seg in 2m 100m
diff --git a/test/t9.tst b/test/t9.tst
index 099462b..a5d2b01 100644
--- a/test/t9.tst
+++ b/test/t9.tst
@@ -59,7 +59,7 @@ do
 	rm -f ${tstf}.pz
 done
 
-for feat in "-B8 -s2m -l1" "-B0 -s2m -l1" "-D -s10k -l1" "-D -F -s2m -l1" "-p -e AES -s2m -l1" "-s2m -l15" "-e AES -k64" "-e SALSA20 -k8" "-e AES -k8" "-e SALSA20 -k64"
+for feat in "-B8 -s2m -l1" "-B-1 -s2m -l1" "-D -s10k -l1" "-D -F -s2m -l1" "-p -e AES -s2m -l1" "-s2m -l15" "-e AES -k64" "-e SALSA20 -k8" "-e AES -k8" "-e SALSA20 -k64"
 do
 	for algo in lzfx lz4 zlib bzip2 libbsc ppmd lzma
 	do