Illumos 5746 - more checksumming in zfs send

5746 more checksumming in zfs send
Reviewed by: Christopher Siden <[email protected]>
Reviewed by: George Wilson <[email protected]>
Reviewed by: Bayard Bell <[email protected]>
Approved by: Albert Lee <[email protected]>

References:
  https://www.illumos.org/issues/5746
  illumos/illumos-gate@98110f0
  #905

Porting notes:
- Minor conflicts due to:
  - 2024041
  - 044baf0
  - 88904bb
- Fix ISO C90 warnings (-Werror=declaration-after-statement)
  - arc_buf_t *abuf;
  - dmu_buf_t *bonus;
  - zio_cksum_t cksum_orig;
  - zio_cksum_t *cksump;
- Fix format '%llx' format specifier warning
- Align message in zstreamdump safe_malloc() with upstream

Ported-by: kernelOfTruth [email protected]
Signed-off-by: Brian Behlendorf <[email protected]>
Closes #3611

Author: ahrens

cmd/zstreamdump/zstreamdump.c

@@ -27,7 +27,7 @@
  */
 
 /*
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2013, 2014 by Delphix. All rights reserved.
  */
 
 #include <ctype.h>
@@ -36,6 +36,7 @@
 #include <stdlib.h>
 #include <strings.h>
 #include <unistd.h>
+#include <stddef.h>
 
 #include <sys/dmu.h>
 #include <sys/zfs_ioctl.h>
@@ -73,8 +74,8 @@ safe_malloc(size_t size)
 {
 	void *rv = malloc(size);
 	if (rv == NULL) {
-		(void) fprintf(stderr, "ERROR; failed to allocate %u bytes\n",
-		    (unsigned)size);
+		(void) fprintf(stderr, "ERROR; failed to allocate %zu bytes\n",
+		    size);
 		abort();
 	}
 	return (rv);
@@ -85,7 +86,6 @@ safe_malloc(size_t size)
  *
  * Read while computing incremental checksum
  */
-
 static size_t
 ssread(void *buf, size_t len, zio_cksum_t *cksum)
 {
@@ -94,7 +94,7 @@ ssread(void *buf, size_t len, zio_cksum_t *cksum)
 	if ((outlen = fread(buf, len, 1, send_stream)) == 0)
 		return (0);
 
-	if (do_cksum && cksum) {
+	if (do_cksum) {
 		if (do_byteswap)
 			fletcher_4_incremental_byteswap(buf, len, cksum);
 		else
@@ -104,6 +104,34 @@ ssread(void *buf, size_t len, zio_cksum_t *cksum)
 	return (outlen);
 }
 
+static size_t
+read_hdr(dmu_replay_record_t *drr, zio_cksum_t *cksum)
+{
+	ASSERT3U(offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum),
+	    ==, sizeof (dmu_replay_record_t) - sizeof (zio_cksum_t));
+	size_t r = ssread(drr, sizeof (*drr) - sizeof (zio_cksum_t), cksum);
+	if (r == 0)
+		return (0);
+	zio_cksum_t saved_cksum = *cksum;
+	r = ssread(&drr->drr_u.drr_checksum.drr_checksum,
+	    sizeof (zio_cksum_t), cksum);
+	if (r == 0)
+		return (0);
+	if (!ZIO_CHECKSUM_IS_ZERO(&drr->drr_u.drr_checksum.drr_checksum) &&
+	    !ZIO_CHECKSUM_EQUAL(saved_cksum,
+	    drr->drr_u.drr_checksum.drr_checksum)) {
+		fprintf(stderr, "invalid checksum\n");
+		(void) printf("Incorrect checksum in record header.\n");
+		(void) printf("Expected checksum = %llx/%llx/%llx/%llx\n",
+		    (longlong_t)saved_cksum.zc_word[0],
+		    (longlong_t)saved_cksum.zc_word[1],
+		    (longlong_t)saved_cksum.zc_word[2],
+		    (longlong_t)saved_cksum.zc_word[3]);
+		exit(1);
+	}
+	return (sizeof (*drr));
+}
+
 /*
  * Print part of a block in ASCII characters
  */
@@ -135,7 +163,7 @@ print_block(char *buf, int length)
 	 * Start printing ASCII characters at a constant offset, after
 	 * the hex prints. Leave 3 characters per byte on a line (2 digit
 	 * hex number plus 1 space) plus spaces between characters and
-	 * groupings
+	 * groupings.
 	 */
 	int ascii_start = BYTES_PER_LINE * 3 +
 	    BYTES_PER_LINE / DUMP_GROUPING + 2;
@@ -185,8 +213,10 @@ main(int argc, char *argv[])
 	struct drr_free *drrf = &thedrr.drr_u.drr_free;
 	struct drr_spill *drrs = &thedrr.drr_u.drr_spill;
 	struct drr_write_embedded *drrwe = &thedrr.drr_u.drr_write_embedded;
+	struct drr_checksum *drrc = &thedrr.drr_u.drr_checksum;
 	char c;
 	boolean_t verbose = B_FALSE;
+	boolean_t very_verbose = B_FALSE;
 	boolean_t first = B_TRUE;
 	/*
 	 * dump flag controls whether the contents of any modified data blocks
@@ -204,11 +234,14 @@ main(int argc, char *argv[])
 			do_cksum = B_FALSE;
 			break;
 		case 'v':
+			if (verbose)
+				very_verbose = B_TRUE;
 			verbose = B_TRUE;
 			break;
 		case 'd':
 			dump = B_TRUE;
 			verbose = B_TRUE;
+			very_verbose = B_TRUE;
 			break;
 		case ':':
 			(void) fprintf(stderr,
@@ -231,7 +264,7 @@ main(int argc, char *argv[])
 	}
 
 	send_stream = stdin;
-	while (ssread(drr, sizeof (dmu_replay_record_t), &zc)) {
+	while (read_hdr(drr, &zc)) {
 
 		/*
 		 * If this is the first DMU record being processed, check for
@@ -437,7 +470,7 @@ main(int argc, char *argv[])
 			if (verbose) {
 				(void) printf("WRITE object = %llu type = %u "
 				    "checksum type = %u\n"
-				    "offset = %llu length = %llu "
+				    "    offset = %llu length = %llu "
 				    "props = %llx\n",
 				    (u_longlong_t)drrw->drr_object,
 				    drrw->drr_type,
@@ -481,9 +514,9 @@ main(int argc, char *argv[])
 			if (verbose) {
 				(void) printf("WRITE_BYREF object = %llu "
 				    "checksum type = %u props = %llx\n"
-				    "offset = %llu length = %llu\n"
+				    "    offset = %llu length = %llu\n"
 				    "toguid = %llx refguid = %llx\n"
-				    "refobject = %llu refoffset = %llu\n",
+				    "    refobject = %llu refoffset = %llu\n",
 				    (u_longlong_t)drrwbr->drr_object,
 				    drrwbr->drr_checksumtype,
 				    (u_longlong_t)drrwbr->drr_key.ddk_prop,
@@ -544,7 +577,7 @@ main(int argc, char *argv[])
 			if (verbose) {
 				(void) printf("WRITE_EMBEDDED object = %llu "
 				    "offset = %llu length = %llu\n"
-				    "toguid = %llx comp = %u etype = %u "
+				    "    toguid = %llx comp = %u etype = %u "
 				    "lsize = %u psize = %u\n",
 				    (u_longlong_t)drrwe->drr_object,
 				    (u_longlong_t)drrwe->drr_offset,
@@ -562,6 +595,13 @@ main(int argc, char *argv[])
 			/* should never be reached */
 			exit(1);
 		}
+		if (drr->drr_type != DRR_BEGIN && very_verbose) {
+			(void) printf("    checksum = %llx/%llx/%llx/%llx\n",
+			    (longlong_t)drrc->drr_checksum.zc_word[0],
+			    (longlong_t)drrc->drr_checksum.zc_word[1],
+			    (longlong_t)drrc->drr_checksum.zc_word[2],
+			    (longlong_t)drrc->drr_checksum.zc_word[3]);
+		}
 		pcksum = zc;
 	}
 	free(buf);

include/sys/spa.h

@@ -446,6 +446,19 @@ _NOTE(CONSTCOND) } while (0)
 	((zc1).zc_word[2] - (zc2).zc_word[2]) | \
 	((zc1).zc_word[3] - (zc2).zc_word[3])))
 
+#define	ZIO_CHECKSUM_IS_ZERO(zc) \
+	(0 == ((zc)->zc_word[0] | (zc)->zc_word[1] | \
+	(zc)->zc_word[2] | (zc)->zc_word[3]))
+
+#define	ZIO_CHECKSUM_BSWAP(zcp)					\
+{								\
+	(zcp)->zc_word[0] = BSWAP_64((zcp)->zc_word[0]);	\
+	(zcp)->zc_word[1] = BSWAP_64((zcp)->zc_word[1]);	\
+	(zcp)->zc_word[2] = BSWAP_64((zcp)->zc_word[2]);	\
+	(zcp)->zc_word[3] = BSWAP_64((zcp)->zc_word[3]);	\
+}
+
+
 #define	DVA_IS_VALID(dva)	(DVA_GET_ASIZE(dva) != 0)
 
 #define	ZIO_SET_CHECKSUM(zcp, w0, w1, w2, w3)	\

include/sys/zfs_ioctl.h

@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  */
 
 #ifndef	_SYS_ZFS_IOCTL_H
@@ -237,6 +237,22 @@ typedef struct dmu_replay_record {
 			uint32_t drr_psize; /* compr. (real) size of payload */
 			/* (possibly compressed) content follows */
 		} drr_write_embedded;
+
+		/*
+		 * Nore: drr_checksum is overlaid with all record types
+		 * except DRR_BEGIN.  Therefore its (non-pad) members
+		 * must not overlap with members from the other structs.
+		 * We accomplish this by putting its members at the very
+		 * end of the struct.
+		 */
+		struct drr_checksum {
+			uint64_t drr_pad[34];
+			/*
+			 * fletcher-4 checksum of everything preceding the
+			 * checksum.
+			 */
+			zio_cksum_t drr_checksum;
+		} drr_checksum;
 	} drr_u;
 } dmu_replay_record_t;
 

include/sys/zio_checksum.h

@@ -20,6 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2014 by Delphix. All rights reserved.
  */
 
 #ifndef _SYS_ZIO_CHECKSUM_H
@@ -34,13 +35,13 @@ extern "C" {
 /*
  * Signature for checksum functions.
  */
-typedef void zio_checksum_t(const void *data, uint64_t size, zio_cksum_t *zcp);
+typedef void zio_checksum_func_t(const void *, uint64_t, zio_cksum_t *);
 
 /*
  * Information about each checksum function.
  */
 typedef const struct zio_checksum_info {
-	zio_checksum_t	*ci_func[2]; /* checksum function for each byteorder */
+	zio_checksum_func_t *ci_func[2]; /* checksum function per byteorder */
 	int		ci_correctable;	/* number of correctable bits	*/
 	int		ci_eck;		/* uses zio embedded checksum? */
 	int		ci_dedup;	/* strong enough for dedup? */
@@ -61,7 +62,7 @@ extern zio_checksum_info_t zio_checksum_table[ZIO_CHECKSUM_FUNCTIONS];
 /*
  * Checksum routines.
  */
-extern zio_checksum_t zio_checksum_SHA256;
+extern zio_checksum_func_t zio_checksum_SHA256;
 
 extern void zio_checksum_compute(zio_t *zio, enum zio_checksum checksum,
     void *data, uint64_t size);