From f3fd0e8a7141ec68e6b20577e8669aeb8b42f6a0 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrik=20Sj=C3=B6lin?=
 <44264371+ulrik-s@users.noreply.github.com>
Date: Tue, 24 Jun 2025 19:25:11 +0200
Subject: [PATCH 01/20] Apply whitespace fixes

---
 Documentation/gitprotocol-capabilities.adoc |   8 +
 Makefile                                    | 139 ++++++-------
 alloc.c                                     |   8 +
 alloc.h                                     |   1 +
 bblob.c                                     | 213 ++++++++++++++++++++
 bblob.h                                     |  25 +++
 builtin/cat-file.c                          |  12 +-
 builtin/fsck.c                              |   6 +-
 builtin/repack.c                            |  38 +++-
 fsck.c                                      | 111 +++++++++-
 fsck.h                                      |  45 +++--
 object-file.c                               |  48 +++--
 object-store.c                              |  37 +++-
 object-store.h                              |  11 +-
 object.c                                    |  68 ++++---
 object.h                                    |   6 +-
 serve.c                                     |   4 +
 streaming.c                                 |  35 +++-
 t/meson.build                               |   7 +-
 t/t1030-bblob.sh                            |  37 ++++
 upload-pack.c                               |   1 +
 21 files changed, 687 insertions(+), 173 deletions(-)
 create mode 100644 bblob.c
 create mode 100644 bblob.h
 create mode 100755 t/t1030-bblob.sh

diff --git a/Documentation/gitprotocol-capabilities.adoc b/Documentation/gitprotocol-capabilities.adoc
index 2cf7735be479e6..c46cbd28e69e1a 100644
--- a/Documentation/gitprotocol-capabilities.adoc
+++ b/Documentation/gitprotocol-capabilities.adoc
@@ -378,6 +378,14 @@ fetch-pack may send "filter" commands to request a partial clone
 or partial fetch and request that the server omit various objects
 from the packfile.
 
+bblob
+-----
+
+If the upload-pack server advertises the 'bblob' capability, it may
+send packfiles containing bblob objects. Clients that do not
+understand this object type must refrain from requesting the
+capability.
+
 session-id=<session-id>
 -----------------------
 
diff --git a/Makefile b/Makefile
index 70d1543b6b8688..fb4a12b76d3fe6 100644
--- a/Makefile
+++ b/Makefile
@@ -974,6 +974,7 @@ LIB_OBJS += base85.o
 LIB_OBJS += bisect.o
 LIB_OBJS += blame.o
 LIB_OBJS += blob.o
+LIB_OBJS += bblob.o
 LIB_OBJS += bloom.o
 LIB_OBJS += branch.o
 LIB_OBJS += bulk-checkin.o
@@ -1466,7 +1467,7 @@ GIT-SPATCH-DEFINES: FORCE
 	    if test x"$$FLAGS" != x"`cat GIT-SPATCH-DEFINES 2>/dev/null`" ; then \
 		echo >&2 "    * new spatch flags"; \
 		echo "$$FLAGS" >GIT-SPATCH-DEFINES; \
-            fi
+	    fi
 
 include config.mak.uname
 -include config.mak.autogen
@@ -1601,23 +1602,23 @@ ifneq (,$(SOCKLEN_T))
 endif
 
 ifeq ($(uname_S),Darwin)
-        ifndef NO_FINK
-                ifeq ($(shell test -d /sw/lib && echo y),y)
+	ifndef NO_FINK
+		ifeq ($(shell test -d /sw/lib && echo y),y)
 			BASIC_CFLAGS += -I/sw/include
 			BASIC_LDFLAGS += -L/sw/lib
-                endif
-        endif
-        ifndef NO_DARWIN_PORTS
-                ifeq ($(shell test -d /opt/local/lib && echo y),y)
+		endif
+	endif
+	ifndef NO_DARWIN_PORTS
+		ifeq ($(shell test -d /opt/local/lib && echo y),y)
 			BASIC_CFLAGS += -I/opt/local/include
 			BASIC_LDFLAGS += -L/opt/local/lib
-                endif
-        endif
-        ifndef NO_APPLE_COMMON_CRYPTO
+		endif
+	endif
+	ifndef NO_APPLE_COMMON_CRYPTO
 		NO_OPENSSL = YesPlease
 		APPLE_COMMON_CRYPTO = YesPlease
 		COMPAT_CFLAGS += -DAPPLE_COMMON_CRYPTO
-        endif
+	endif
 	PTHREAD_LIBS =
 endif
 
@@ -1656,23 +1657,23 @@ ifdef NO_CURL
 	REMOTE_CURL_NAMES =
 	EXCLUDED_PROGRAMS += git-http-fetch git-http-push
 else
-        ifdef CURLDIR
+	ifdef CURLDIR
 		# Try "-Wl,-rpath=$(CURLDIR)/$(lib)" in such a case.
 		CURL_CFLAGS = -I$(CURLDIR)/include
 		CURL_LIBCURL = $(call libpath_template,$(CURLDIR)/$(lib))
-        else
+	else
 		CURL_CFLAGS =
 		CURL_LIBCURL =
-        endif
+	endif
 
-        ifndef CURL_LDFLAGS
+	ifndef CURL_LDFLAGS
 		CURL_LDFLAGS = $(eval CURL_LDFLAGS := $$(shell $$(CURL_CONFIG) --libs))$(CURL_LDFLAGS)
-        endif
+	endif
 	CURL_LIBCURL += $(CURL_LDFLAGS)
 
-        ifndef CURL_CFLAGS
+	ifndef CURL_CFLAGS
 		CURL_CFLAGS = $(eval CURL_CFLAGS := $$(shell $$(CURL_CONFIG) --cflags))$(CURL_CFLAGS)
-        endif
+	endif
 	BASIC_CFLAGS += $(CURL_CFLAGS)
 
 	REMOTE_CURL_PRIMARY = git-remote-http$X
@@ -1680,58 +1681,58 @@ else
 	REMOTE_CURL_NAMES = $(REMOTE_CURL_PRIMARY) $(REMOTE_CURL_ALIASES)
 	PROGRAM_OBJS += http-fetch.o
 	PROGRAMS += $(REMOTE_CURL_NAMES)
-        ifndef NO_EXPAT
+	ifndef NO_EXPAT
 		PROGRAM_OBJS += http-push.o
-        endif
+	endif
 	curl_check := $(shell (echo 072200; $(CURL_CONFIG) --vernum | sed -e '/^70[BC]/s/^/0/') 2>/dev/null | sort -r | sed -ne 2p)
-        ifeq "$(curl_check)" "072200"
+	ifeq "$(curl_check)" "072200"
 		USE_CURL_FOR_IMAP_SEND = YesPlease
-        endif
-        ifdef USE_CURL_FOR_IMAP_SEND
+	endif
+	ifdef USE_CURL_FOR_IMAP_SEND
 		BASIC_CFLAGS += -DUSE_CURL_FOR_IMAP_SEND
 		IMAP_SEND_BUILDDEPS = http.o
 		IMAP_SEND_LDFLAGS += $(CURL_LIBCURL)
-        endif
-        ifndef NO_EXPAT
-                ifdef EXPATDIR
+	endif
+	ifndef NO_EXPAT
+		ifdef EXPATDIR
 			BASIC_CFLAGS += -I$(EXPATDIR)/include
 			EXPAT_LIBEXPAT = $(call libpath_template,$(EXPATDIR)/$(lib)) -lexpat
-                else
+		else
 			EXPAT_LIBEXPAT = -lexpat
-                endif
-                ifdef EXPAT_NEEDS_XMLPARSE_H
+		endif
+		ifdef EXPAT_NEEDS_XMLPARSE_H
 			BASIC_CFLAGS += -DEXPAT_NEEDS_XMLPARSE_H
-                endif
-        endif
+		endif
+	endif
 endif
 IMAP_SEND_LDFLAGS += $(OPENSSL_LINK) $(OPENSSL_LIBSSL) $(LIB_4_CRYPTO)
 
 ifdef ZLIB_NG
 	BASIC_CFLAGS += -DHAVE_ZLIB_NG
-        ifdef ZLIB_NG_PATH
+	ifdef ZLIB_NG_PATH
 		BASIC_CFLAGS += -I$(ZLIB_NG_PATH)/include
 		EXTLIBS += $(call libpath_template,$(ZLIB_NG_PATH)/$(lib))
-        endif
+	endif
 	EXTLIBS += -lz-ng
 else
-        ifdef ZLIB_PATH
+	ifdef ZLIB_PATH
 		BASIC_CFLAGS += -I$(ZLIB_PATH)/include
 		EXTLIBS += $(call libpath_template,$(ZLIB_PATH)/$(lib))
-        endif
+	endif
 	EXTLIBS += -lz
 endif
 
 ifndef NO_OPENSSL
 	OPENSSL_LIBSSL = -lssl
-        ifdef OPENSSLDIR
+	ifdef OPENSSLDIR
 		BASIC_CFLAGS += -I$(OPENSSLDIR)/include
 		OPENSSL_LINK = $(call libpath_template,$(OPENSSLDIR)/$(lib))
-        else
+	else
 		OPENSSL_LINK =
-        endif
-        ifdef NEEDS_CRYPTO_WITH_SSL
+	endif
+	ifdef NEEDS_CRYPTO_WITH_SSL
 		OPENSSL_LIBSSL += -lcrypto
-        endif
+	endif
 else
 	BASIC_CFLAGS += -DNO_OPENSSL
 	OPENSSL_LIBSSL =
@@ -1749,18 +1750,18 @@ ifdef APPLE_COMMON_CRYPTO
 endif
 endif
 ifndef NO_ICONV
-        ifdef NEEDS_LIBICONV
-                ifdef ICONVDIR
+	ifdef NEEDS_LIBICONV
+		ifdef ICONVDIR
 			BASIC_CFLAGS += -I$(ICONVDIR)/include
 			ICONV_LINK = $(call libpath_template,$(ICONVDIR)/$(lib))
-                else
+		else
 			ICONV_LINK =
-                endif
-                ifdef NEEDS_LIBINTL_BEFORE_LIBICONV
+		endif
+		ifdef NEEDS_LIBINTL_BEFORE_LIBICONV
 			ICONV_LINK += -lintl
-                endif
+		endif
 		EXTLIBS += $(ICONV_LINK) -liconv
-        endif
+	endif
 endif
 ifdef ICONV_OMITS_BOM
 	BASIC_CFLAGS += -DICONV_OMITS_BOM
@@ -1880,10 +1881,10 @@ ifdef NO_MMAP
 	COMPAT_CFLAGS += -DNO_MMAP
 	COMPAT_OBJS += compat/mmap.o
 else
-        ifdef USE_WIN32_MMAP
+	ifdef USE_WIN32_MMAP
 		COMPAT_CFLAGS += -DUSE_WIN32_MMAP
 		COMPAT_OBJS += compat/win32mmap.o
-        endif
+	endif
 endif
 ifdef MMAP_PREVENTS_DELETE
 	BASIC_CFLAGS += -DMMAP_PREVENTS_DELETE
@@ -2008,11 +2009,11 @@ else
 	BASIC_CFLAGS += -DSHA1_DC
 	LIB_OBJS += sha1dc_git.o
 ifdef DC_SHA1_EXTERNAL
-        ifdef DC_SHA1_SUBMODULE
-                ifneq ($(DC_SHA1_SUBMODULE),auto)
+	ifdef DC_SHA1_SUBMODULE
+		ifneq ($(DC_SHA1_SUBMODULE),auto)
 $(error Only set DC_SHA1_EXTERNAL or DC_SHA1_SUBMODULE, not both)
-                endif
-        endif
+		endif
+	endif
 	BASIC_CFLAGS += -DDC_SHA1_EXTERNAL
 	EXTLIBS += -lsha1detectcoll
 else
@@ -2218,26 +2219,26 @@ endif
 
 ifdef RUNTIME_PREFIX
 
-        ifdef HAVE_BSD_KERN_PROC_SYSCTL
+	ifdef HAVE_BSD_KERN_PROC_SYSCTL
 		BASIC_CFLAGS += -DHAVE_BSD_KERN_PROC_SYSCTL
-        endif
+	endif
 
-        ifneq ($(PROCFS_EXECUTABLE_PATH),)
+	ifneq ($(PROCFS_EXECUTABLE_PATH),)
 		pep_SQ = $(subst ','\'',$(PROCFS_EXECUTABLE_PATH))
 		BASIC_CFLAGS += '-DPROCFS_EXECUTABLE_PATH="$(pep_SQ)"'
-        endif
+	endif
 
-        ifdef HAVE_NS_GET_EXECUTABLE_PATH
+	ifdef HAVE_NS_GET_EXECUTABLE_PATH
 		BASIC_CFLAGS += -DHAVE_NS_GET_EXECUTABLE_PATH
-        endif
+	endif
 
-        ifdef HAVE_ZOS_GET_EXECUTABLE_PATH
+	ifdef HAVE_ZOS_GET_EXECUTABLE_PATH
 		BASIC_CFLAGS += -DHAVE_ZOS_GET_EXECUTABLE_PATH
-        endif
+	endif
 
-        ifdef HAVE_WPGMPTR
+	ifdef HAVE_WPGMPTR
 		BASIC_CFLAGS += -DHAVE_WPGMPTR
-        endif
+	endif
 
 endif
 
@@ -2595,7 +2596,7 @@ GIT-SCRIPT-DEFINES: FORCE
 	    if test x"$$FLAGS" != x"`cat $@ 2>/dev/null`" ; then \
 		echo >&2 "    * new script parameters"; \
 		echo "$$FLAGS" >$@; \
-            fi
+	    fi
 
 $(SCRIPT_SH_GEN) $(SCRIPT_LIB) : % : %.sh generate-script.sh GIT-BUILD-OPTIONS GIT-SCRIPT-DEFINES
 	$(QUIET_GEN)./generate-script.sh "$<" "$@+" ./GIT-BUILD-OPTIONS && \
@@ -2936,7 +2937,7 @@ Documentation/GIT-EXCLUDED-PROGRAMS: FORCE
 		x"`cat Documentation/GIT-EXCLUDED-PROGRAMS 2>/dev/null`" ; then \
 		echo >&2 "    * new documentation flags"; \
 		echo "$$EXCLUDED" >Documentation/GIT-EXCLUDED-PROGRAMS; \
-            fi
+	    fi
 
 .PHONY: doc man man-perl html info pdf
 doc: man-perl
@@ -3172,7 +3173,7 @@ GIT-CFLAGS: FORCE
 	    if test x"$$FLAGS" != x"`cat GIT-CFLAGS 2>/dev/null`" ; then \
 		echo >&2 "    * new build flags"; \
 		echo "$$FLAGS" >GIT-CFLAGS; \
-            fi
+	    fi
 
 TRACK_LDFLAGS = $(subst ','\'',$(ALL_LDFLAGS))
 
@@ -3181,7 +3182,7 @@ GIT-LDFLAGS: FORCE
 	    if test x"$$FLAGS" != x"`cat GIT-LDFLAGS 2>/dev/null`" ; then \
 		echo >&2 "    * new link flags"; \
 		echo "$$FLAGS" >GIT-LDFLAGS; \
-            fi
+	    fi
 
 ifdef RUNTIME_PREFIX
 RUNTIME_PREFIX_OPTION = true
@@ -3256,7 +3257,7 @@ GIT-PYTHON-VARS: FORCE
 	    if test x"$$VARS" != x"`cat $@ 2>/dev/null`" ; then \
 		echo >&2 "    * new Python interpreter location"; \
 		echo "$$VARS" >$@; \
-            fi
+	    fi
 endif
 
 test_bindir_programs := $(patsubst %,bin-wrappers/%,$(BINDIR_PROGRAMS_NEED_X) $(BINDIR_PROGRAMS_NO_X) $(TEST_PROGRAMS_NEED_X))
@@ -3927,7 +3928,7 @@ GIT-TEST-SUITES: FORCE
 	    if test x"$$FLAGS" != x"`cat GIT-TEST-SUITES 2>/dev/null`" ; then \
 		echo >&2 "    * new test suites"; \
 		echo "$$FLAGS" >GIT-TEST-SUITES; \
-            fi
+	    fi
 
 $(UNIT_TEST_DIR)/clar-decls.h: $(patsubst %,$(UNIT_TEST_DIR)/%.c,$(CLAR_TEST_SUITES)) $(UNIT_TEST_DIR)/generate-clar-decls.sh GIT-TEST-SUITES
 	$(QUIET_GEN)$(SHELL_PATH) $(UNIT_TEST_DIR)/generate-clar-decls.sh "$@" $(filter %.c,$^)
diff --git a/alloc.c b/alloc.c
index 377e80f5dda2f8..bb87ead8e1ca43 100644
--- a/alloc.c
+++ b/alloc.c
@@ -11,6 +11,7 @@
 #include "git-compat-util.h"
 #include "object.h"
 #include "blob.h"
+#include "bblob.h"
 #include "tree.h"
 #include "commit.h"
 #include "repository.h"
@@ -77,6 +78,13 @@ void *alloc_blob_node(struct repository *r)
 	return b;
 }
 
+void *alloc_bblob_node(struct repository *r)
+{
+       struct bblob *bb = alloc_node(r->parsed_objects->blob_state, sizeof(struct bblob));
+       bb->object.type = OBJ_BBLOB;
+       return bb;
+}
+
 void *alloc_tree_node(struct repository *r)
 {
 	struct tree *t = alloc_node(r->parsed_objects->tree_state, sizeof(struct tree));
diff --git a/alloc.h b/alloc.h
index 3f4a0ad310a94b..662205812928f2 100644
--- a/alloc.h
+++ b/alloc.h
@@ -8,6 +8,7 @@ struct tag;
 struct repository;
 
 void *alloc_blob_node(struct repository *r);
+void *alloc_bblob_node(struct repository *r);
 void *alloc_tree_node(struct repository *r);
 void init_commit_node(struct commit *c);
 void *alloc_commit_node(struct repository *r);
diff --git a/bblob.c b/bblob.c
new file mode 100644
index 00000000000000..4d2910495548bb
--- /dev/null
+++ b/bblob.c
@@ -0,0 +1,213 @@
+#include "git-compat-util.h"
+#include "bblob.h"
+#include "alloc.h"
+#include "object-file.h"
+#include "hash.h"
+#include "repository.h"
+#include "object-store.h"
+#include "streaming.h"
+
+extern int disable_bblob_conversion;
+
+const char *bblob_type = "bblob";
+
+struct bblob *lookup_bblob(struct repository *r, const struct object_id *oid)
+{
+       struct object *obj = lookup_object(r, oid);
+       if (!obj)
+	       return create_object(r, oid, alloc_bblob_node(r));
+       return object_as_type(obj, OBJ_BBLOB, 0);
+}
+
+void parse_bblob_buffer(struct bblob *item)
+{
+	item->object.parsed = 1;
+}
+
+static int write_bblob_tree(struct repository *r, struct object_id *oids,
+			   int nr, struct object_id *oid)
+{
+       size_t oidsz = r->hash_algo->rawsz;
+       if (nr <= BBLOB_FANOUT) {
+	       size_t rawlen = oidsz * BBLOB_FANOUT;
+	       void *raw = xcalloc(1, rawlen);
+	       for (int i = 0; i < nr; i++)
+		       memcpy((char *)raw + i * oidsz, oids[i].hash, oidsz);
+	       int ret = write_object_file(raw, rawlen, OBJ_BBLOB, oid);
+	       free(raw);
+	       return ret;
+       }
+
+       int groups = (nr + BBLOB_FANOUT - 1) / BBLOB_FANOUT;
+       struct object_id *tmp = xcalloc(groups, sizeof(*tmp));
+       for (int i = 0; i < groups; i++) {
+	       int this = nr - i * BBLOB_FANOUT;
+	       if (this > BBLOB_FANOUT)
+		       this = BBLOB_FANOUT;
+	       if (write_bblob_tree(r, oids + i * BBLOB_FANOUT, this, &tmp[i])) {
+		       free(tmp);
+		       return -1;
+	       }
+       }
+       int ret = write_bblob_tree(r, tmp, groups, oid);
+       free(tmp);
+       return ret;
+}
+
+int write_bblob(struct repository *r, const void *buf, unsigned long len,
+	       struct object_id *oid)
+{
+       size_t oids_alloc = 0, oids_nr = 0;
+       struct object_id *oids = NULL;
+       unsigned char window[64];
+       size_t win_len = 0;
+       size_t chunk_start = 0;
+
+       for (size_t i = 0; i < len; i++) {
+	       window[win_len % 64] = ((const unsigned char *)buf)[i];
+	       if (win_len >= 63 && i - chunk_start + 1 >= BBLOB_CHUNK_GOAL) {
+		       struct git_hash_ctx c;
+		       unsigned char out[GIT_MAX_RAWSZ];
+		       r->hash_algo->init_fn(&c);
+		       git_hash_update(&c, window, 64);
+		       git_hash_final(out, &c);
+		       unsigned short bits =
+			       (out[r->hash_algo->rawsz - 2] << 8) |
+			       out[r->hash_algo->rawsz - 1];
+		       if ((bits & 0x1fff) == 0) {
+			       struct object_id ch;
+			       disable_bblob_conversion++;
+			       if (write_object_file((const char *)buf + chunk_start,
+						    i - chunk_start + 1,
+						    OBJ_BLOB, &ch)) {
+				       disable_bblob_conversion--;
+				       free(oids);
+				       return -1;
+			       }
+			       disable_bblob_conversion--;
+			       ALLOC_GROW(oids, oids_nr + 1, oids_alloc);
+			       oidcpy(&oids[oids_nr++], &ch);
+			       chunk_start = i + 1;
+		       }
+	       }
+	       win_len++;
+       }
+       if (chunk_start < len) {
+	       struct object_id ch;
+	       disable_bblob_conversion++;
+	       if (write_object_file((const char *)buf + chunk_start,
+				    len - chunk_start,
+				    OBJ_BLOB, &ch)) {
+		       disable_bblob_conversion--;
+		       free(oids);
+		       return -1;
+	       }
+	       disable_bblob_conversion--;
+	       ALLOC_GROW(oids, oids_nr + 1, oids_alloc);
+	       oidcpy(&oids[oids_nr++], &ch);
+       }
+
+       int ret = write_bblob_tree(r, oids, oids_nr, oid);
+       free(oids);
+       return ret;
+}
+
+static void *read_raw(struct repository *r, const struct object_id *oid,
+		      enum object_type *type, unsigned long *size)
+{
+       struct object_info oi = OBJECT_INFO_INIT;
+       void *data;
+
+       oi.typep = type;
+       oi.sizep = size;
+       oi.contentp = &data;
+       if (oid_object_info_extended(r, oid, &oi,
+				    OBJECT_INFO_DIE_IF_CORRUPT | OBJECT_INFO_LOOKUP_REPLACE))
+	       return NULL;
+       return data;
+}
+
+static void *read_bblob_rec(struct repository *r, const struct object_id *oid,
+			    unsigned long *size)
+{
+       enum object_type t;
+       unsigned long sz;
+       void *data = read_raw(r, oid, &t, &sz);
+       if (!data)
+	       return NULL;
+       if (t == OBJ_BLOB) {
+	       *size = sz;
+	       return data;
+       }
+       if (t != OBJ_BBLOB) {
+	       free(data);
+	       return NULL;
+       }
+
+       size_t oidsz = r->hash_algo->rawsz;
+       int cnt = sz / oidsz;
+       unsigned long out_sz = 0;
+       char *out = NULL;
+       for (int i = 0; i < cnt; i++) {
+	       struct object_id child;
+	       memset(&child, 0, sizeof(child));
+	       memcpy(child.hash, (char *)data + i * oidsz, oidsz);
+	       if (is_null_oid(&child))
+		       continue;
+	       unsigned long csz;
+	       void *cbuf = read_bblob_rec(r, &child, &csz);
+	       if (!cbuf) {
+		       free(out);
+		       free(data);
+		       return NULL;
+	       }
+	       REALLOC_ARRAY(out, out_sz + csz);
+	       memcpy(out + out_sz, cbuf, csz);
+	       out_sz += csz;
+	       free(cbuf);
+       }
+       free(data);
+       *size = out_sz;
+       return out;
+}
+
+static unsigned long size_bblob_rec(struct repository *r, const struct object_id *oid)
+{
+       enum object_type t;
+       unsigned long sz;
+       void *data = read_raw(r, oid, &t, &sz);
+       if (!data)
+	       return 0;
+       if (t == OBJ_BLOB) {
+	       free(data);
+	       return sz;
+       }
+       if (t != OBJ_BBLOB) {
+	       free(data);
+	       return 0;
+       }
+       size_t oidsz = r->hash_algo->rawsz;
+       int cnt = sz / oidsz;
+       unsigned long total = 0;
+       for (int i = 0; i < cnt; i++) {
+	       struct object_id child;
+	       memset(&child, 0, sizeof(child));
+	       memcpy(child.hash, (char *)data + i * oidsz, oidsz);
+	       if (is_null_oid(&child))
+		       continue;
+	       total += size_bblob_rec(r, &child);
+       }
+       free(data);
+       return total;
+}
+
+void *read_bblob(struct repository *r, const struct object_id *oid,
+		 unsigned long *size)
+{
+       return read_bblob_rec(r, oid, size);
+}
+
+unsigned long bblob_size(struct repository *r, const struct object_id *oid)
+{
+       return size_bblob_rec(r, oid);
+}
diff --git a/bblob.h b/bblob.h
new file mode 100644
index 00000000000000..370f6461c62205
--- /dev/null
+++ b/bblob.h
@@ -0,0 +1,25 @@
+#ifndef BBLOB_H
+#define BBLOB_H
+
+#include "object.h"
+
+/* Number of child entries in each bblob node */
+#define BBLOB_FANOUT 64
+
+/* heuristic target chunk size when splitting large blobs */
+#define BBLOB_CHUNK_GOAL 4096
+
+struct bblob {
+       struct object object;
+       struct object_id oids[BBLOB_FANOUT];
+};
+
+struct bblob *lookup_bblob(struct repository *r, const struct object_id *oid);
+void parse_bblob_buffer(struct bblob *item);
+int write_bblob(struct repository *r, const void *buf, unsigned long len,
+	       struct object_id *oid);
+void *read_bblob(struct repository *r, const struct object_id *oid,
+		unsigned long *size);
+unsigned long bblob_size(struct repository *r, const struct object_id *oid);
+
+#endif /* BBLOB_H */
diff --git a/builtin/cat-file.c b/builtin/cat-file.c
index 67a5ff2b9ebd29..3bf27cb315aa66 100644
--- a/builtin/cat-file.c
+++ b/builtin/cat-file.c
@@ -25,6 +25,7 @@
 #include "object-file.h"
 #include "object-name.h"
 #include "object-store.h"
+#include "bblob.h"
 #include "replace-object.h"
 #include "promisor-remote.h"
 #include "mailmap.h"
@@ -149,7 +150,16 @@ static int cat_one_file(int opt, const char *exp_type, const char *obj_name)
 		if (oid_object_info_extended(the_repository, &oid, &oi, flags) < 0)
 			die("git cat-file: could not get object info");
 
-		if (use_mailmap && (type == OBJ_COMMIT || type == OBJ_TAG)) {
+		if (!oi.typep) {
+			struct object_info ti = OBJECT_INFO_INIT;
+			ti.typep = &type;
+			if (oid_object_info_extended(the_repository, &oid, &ti, flags) < 0)
+				die("git cat-file: could not get object info");
+		}
+
+		if (type == OBJ_BBLOB)
+			size = bblob_size(the_repository, &oid);
+		else if (use_mailmap && (type == OBJ_COMMIT || type == OBJ_TAG)) {
 			size_t s = size;
 			buf = replace_idents_using_mailmap(buf, &s);
 			size = cast_size_t_to_ulong(s);
diff --git a/builtin/fsck.c b/builtin/fsck.c
index e7d96a9c8ea586..81f9bd9535c0d8 100644
--- a/builtin/fsck.c
+++ b/builtin/fsck.c
@@ -946,8 +946,10 @@ static struct option fsck_opts[] = {
 				N_("write dangling objects in .git/lost-found")),
 	OPT_BOOL(0, "progress", &show_progress, N_("show progress")),
 	OPT_BOOL(0, "name-objects", &name_objects, N_("show verbose names for reachable objects")),
-	OPT_BOOL(0, "references", &check_references, N_("check reference database consistency")),
-	OPT_END(),
+	   OPT_BOOL(0, "references", &check_references, N_("check reference database consistency")),
+	   OPT_BOOL(0, "full-bblob-verify", &fsck_obj_options.full_bblob_verify,
+		N_("reassemble bblobs to verify contents")),
+	   OPT_END(),
 };
 
 int cmd_fsck(int argc,
diff --git a/builtin/repack.c b/builtin/repack.c
index 59214dbdfdfcd5..908f709e69342d 100644
--- a/builtin/repack.c
+++ b/builtin/repack.c
@@ -18,6 +18,7 @@
 #include "packfile.h"
 #include "prune-packed.h"
 #include "object-store.h"
+#include "bblob.h"
 #include "promisor-remote.h"
 #include "shallow.h"
 #include "pack.h"
@@ -38,8 +39,30 @@ static int pack_kept_objects = -1;
 static int write_bitmaps = -1;
 static int use_delta_islands;
 static int run_update_server_info = 1;
+static int convert_to_bblob;
 static char *packdir, *packtmp_name, *packtmp;
 
+static int convert_one_loose(const struct object_id *oid, const char *path,
+			    void *data)
+{
+	   enum object_type type;
+	   unsigned long size;
+	   void *buf = repo_read_object_file(the_repository, oid, &type, &size);
+	   if (!buf)
+	       return 0;
+	   if (type == OBJ_BLOB) {
+	       struct object_id new_oid;
+	       write_bblob(the_repository, buf, size, &new_oid);
+	   }
+	   free(buf);
+	   return 0;
+}
+
+static void convert_all_blobs_to_bblob(void)
+{
+	   for_each_loose_object(convert_one_loose, NULL, 0);
+}
+
 static const char *const git_repack_usage[] = {
 	N_("git repack [-a] [-A] [-d] [-f] [-F] [-l] [-n] [-q] [-b] [-m]\n"
 	   "[--window=<n>] [--depth=<n>] [--threads=<n>] [--keep-pack=<pack-name>]\n"
@@ -1207,9 +1230,11 @@ int cmd_repack(int argc,
 				N_("limits the maximum number of threads")),
 		OPT_UNSIGNED(0, "max-pack-size", &po_args.max_pack_size,
 			     N_("maximum size of each packfile")),
-		OPT_PARSE_LIST_OBJECTS_FILTER(&po_args.filter_options),
-		OPT_BOOL(0, "pack-kept-objects", &pack_kept_objects,
-				N_("repack objects in packs marked with .keep")),
+	       OPT_PARSE_LIST_OBJECTS_FILTER(&po_args.filter_options),
+	       OPT_BOOL(0, "convert-to-bblob", &convert_to_bblob,
+			       N_("rewrite all blobs as bblobs")),
+	       OPT_BOOL(0, "pack-kept-objects", &pack_kept_objects,
+			       N_("repack objects in packs marked with .keep")),
 		OPT_STRING_LIST(0, "keep-pack", &keep_pack_list, N_("name"),
 				N_("do not repack this pack")),
 		OPT_INTEGER('g', "geometric", &geometry.split_factor,
@@ -1232,8 +1257,11 @@ int cmd_repack(int argc,
 
 	po_args.window = xstrdup_or_null(opt_window);
 	po_args.window_memory = xstrdup_or_null(opt_window_memory);
-	po_args.depth = xstrdup_or_null(opt_depth);
-	po_args.threads = xstrdup_or_null(opt_threads);
+	   po_args.depth = xstrdup_or_null(opt_depth);
+	   po_args.threads = xstrdup_or_null(opt_threads);
+
+	   if (convert_to_bblob)
+	       convert_all_blobs_to_bblob();
 
 	if (delete_redundant && repository_format_precious_objects)
 		die(_("cannot delete packs in a precious-objects repo"));
diff --git a/fsck.c b/fsck.c
index 8dc8472ceb3781..81eb3dd9382114 100644
--- a/fsck.c
+++ b/fsck.c
@@ -10,6 +10,7 @@
 #include "object.h"
 #include "attr.h"
 #include "blob.h"
+#include "bblob.h"
 #include "tree.h"
 #include "tree-walk.h"
 #include "commit.h"
@@ -480,6 +481,45 @@ static int fsck_walk_tag(struct tag *tag, void *data, struct fsck_options *optio
 	return options->walk(tag->tagged, OBJ_ANY, data, options);
 }
 
+static int fsck_walk_bblob(struct bblob *bb, void *data,
+			   struct fsck_options *options)
+{
+	   enum object_type type;
+	   unsigned long size;
+	   void *buf;
+	   int res = 0;
+	   const char *name = fsck_get_object_name(options, &bb->object.oid);
+
+	   buf = repo_read_raw_object_file(the_repository, &bb->object.oid, &type, &size);
+	   if (!buf || type != OBJ_BBLOB)
+	       return -1;
+
+	   size_t oidsz = the_repository->hash_algo->rawsz;
+	   int cnt = size / oidsz;
+	   for (int i = 0; i < cnt; i++) {
+	       struct object_id child;
+	       memset(&child, 0, sizeof(child));
+	       memcpy(child.hash, (char *)buf + i * oidsz, oidsz);
+	       if (is_null_oid(&child))
+		       continue;
+	       enum object_type ct = oid_object_info(the_repository, &child, NULL);
+	       if (ct <= 0)
+		       continue;
+	       struct object *obj = lookup_object_by_type(the_repository, &child, ct);
+	       if (name && obj)
+		       fsck_put_object_name(options, &child, "%s#%d", name, i);
+	       int result = options->walk(obj, ct, data, options);
+	       if (result < 0) {
+		       free(buf);
+		       return result;
+	       }
+	       if (!res)
+		       res = result;
+	   }
+	   free(buf);
+	   return res;
+}
+
 int fsck_walk(struct object *obj, void *data, struct fsck_options *options)
 {
 	if (!obj)
@@ -488,11 +528,13 @@ int fsck_walk(struct object *obj, void *data, struct fsck_options *options)
 	if (obj->type == OBJ_NONE)
 		parse_object(the_repository, &obj->oid);
 
-	switch (obj->type) {
-	case OBJ_BLOB:
-		return 0;
-	case OBJ_TREE:
-		return fsck_walk_tree((struct tree *)obj, data, options);
+	   switch (obj->type) {
+	   case OBJ_BLOB:
+	       return 0;
+	   case OBJ_BBLOB:
+	       return fsck_walk_bblob((struct bblob *)obj, data, options);
+	   case OBJ_TREE:
+	       return fsck_walk_tree((struct tree *)obj, data, options);
 	case OBJ_COMMIT:
 		return fsck_walk_commit((struct commit *)obj, data, options);
 	case OBJ_TAG:
@@ -1202,7 +1244,52 @@ static int fsck_blob(const struct object_id *oid, const char *buf,
 		}
 	}
 
-	return ret;
+	   return ret;
+}
+
+static int fsck_bblob(const struct object_id *oid, const char *buf,
+		     unsigned long size, struct fsck_options *options)
+{
+	   int ret = 0;
+	   size_t oidsz = the_repository->hash_algo->rawsz;
+
+	   if (size != oidsz * BBLOB_FANOUT)
+	       ret |= report(options, oid, OBJ_BBLOB, FSCK_MSG_BAD_TYPE,
+			      "invalid bblob size");
+
+	   for (int i = 0; i < BBLOB_FANOUT && i * oidsz < size; i++) {
+	       struct object_id child;
+	       memset(&child, 0, sizeof(child));
+	       memcpy(child.hash, buf + i * oidsz, oidsz);
+	       if (is_null_oid(&child))
+		       continue;
+
+	       enum object_type t = oid_object_info(the_repository, &child, NULL);
+	       if (t <= 0) {
+		       ret |= report(options, oid, OBJ_BBLOB,
+				      FSCK_MSG_BAD_OBJECT_SHA1,
+				      "missing child object");
+		       continue;
+	       }
+	       if (t != OBJ_BLOB && t != OBJ_BBLOB)
+		       ret |= report(options, oid, OBJ_BBLOB, FSCK_MSG_BAD_TYPE,
+				      "child has invalid type");
+	       else if (t == OBJ_BBLOB) {
+		       unsigned long csz;
+		       void *cbuf = repo_read_raw_object_file(the_repository,
+							 &child, &t, &csz);
+		       if (!cbuf) {
+			       ret |= report(options, oid, OBJ_BBLOB,
+					      FSCK_MSG_BAD_OBJECT_SHA1,
+					      "cannot read child");
+			       continue;
+		       }
+		       ret |= fsck_bblob(&child, cbuf, csz, options);
+		       free(cbuf);
+	       }
+	   }
+
+	   return ret;
 }
 
 int fsck_object(struct object *obj, void *data, unsigned long size,
@@ -1218,10 +1305,12 @@ int fsck_buffer(const struct object_id *oid, enum object_type type,
 		const void *data, unsigned long size,
 		struct fsck_options *options)
 {
-	if (type == OBJ_BLOB)
-		return fsck_blob(oid, data, size, options);
-	if (type == OBJ_TREE)
-		return fsck_tree(oid, data, size, options);
+	   if (type == OBJ_BLOB)
+	       return fsck_blob(oid, data, size, options);
+	   if (type == OBJ_BBLOB)
+	       return fsck_bblob(oid, data, size, options);
+	   if (type == OBJ_TREE)
+	       return fsck_tree(oid, data, size, options);
 	if (type == OBJ_COMMIT)
 		return fsck_commit(oid, data, size, options);
 	if (type == OBJ_TAG)
@@ -1293,7 +1382,7 @@ static int fsck_blobs(struct oidset *blobs_found, struct oidset *blobs_done,
 		if (oidset_contains(blobs_done, oid))
 			continue;
 
-		buf = repo_read_object_file(the_repository, oid, &type, &size);
+	       buf = repo_read_raw_object_file(the_repository, oid, &type, &size);
 		if (!buf) {
 			if (is_promisor_object(the_repository, oid))
 				continue;
diff --git a/fsck.h b/fsck.h
index 0c5869ac34e216..fc744b5abedb7b 100644
--- a/fsck.h
+++ b/fsck.h
@@ -174,35 +174,40 @@ struct fsck_options {
 	struct oidset gitmodules_done;
 	struct oidset gitattributes_found;
 	struct oidset gitattributes_done;
-	kh_oid_map_t *object_names;
+       kh_oid_map_t *object_names;
+       int full_bblob_verify;
 };
 
 #define FSCK_OPTIONS_DEFAULT { \
-	.skip_oids = OIDSET_INIT, \
-	.gitmodules_found = OIDSET_INIT, \
-	.gitmodules_done = OIDSET_INIT, \
-	.gitattributes_found = OIDSET_INIT, \
-	.gitattributes_done = OIDSET_INIT, \
-	.error_func = fsck_objects_error_function \
+    .skip_oids = OIDSET_INIT, \
+    .gitmodules_found = OIDSET_INIT, \
+    .gitmodules_done = OIDSET_INIT, \
+    .gitattributes_found = OIDSET_INIT, \
+    .gitattributes_done = OIDSET_INIT, \
+    .error_func = fsck_objects_error_function, \
+    .full_bblob_verify = 0 \
 }
 #define FSCK_OPTIONS_STRICT { \
-	.strict = 1, \
-	.gitmodules_found = OIDSET_INIT, \
-	.gitmodules_done = OIDSET_INIT, \
-	.gitattributes_found = OIDSET_INIT, \
-	.gitattributes_done = OIDSET_INIT, \
-	.error_func = fsck_objects_error_function, \
+    .strict = 1, \
+    .gitmodules_found = OIDSET_INIT, \
+    .gitmodules_done = OIDSET_INIT, \
+    .gitattributes_found = OIDSET_INIT, \
+    .gitattributes_done = OIDSET_INIT, \
+    .error_func = fsck_objects_error_function, \
+    .full_bblob_verify = 0, \
 }
 #define FSCK_OPTIONS_MISSING_GITMODULES { \
-	.strict = 1, \
-	.gitmodules_found = OIDSET_INIT, \
-	.gitmodules_done = OIDSET_INIT, \
-	.gitattributes_found = OIDSET_INIT, \
-	.gitattributes_done = OIDSET_INIT, \
-	.error_func = fsck_objects_error_cb_print_missing_gitmodules, \
+    .strict = 1, \
+    .gitmodules_found = OIDSET_INIT, \
+    .gitmodules_done = OIDSET_INIT, \
+    .gitattributes_found = OIDSET_INIT, \
+    .gitattributes_done = OIDSET_INIT, \
+    .error_func = fsck_objects_error_cb_print_missing_gitmodules, \
+    .full_bblob_verify = 0, \
 }
 #define FSCK_REFS_OPTIONS_DEFAULT { \
-	.error_func = fsck_refs_error_function, \
+    .error_func = fsck_refs_error_function, \
+    .full_bblob_verify = 0, \
 }
 
 /* descend in all linked child objects
diff --git a/object-file.c b/object-file.c
index 1ac04c2891634a..288f3dcebf11a7 100644
--- a/object-file.c
+++ b/object-file.c
@@ -21,6 +21,7 @@
 #include "loose.h"
 #include "object-file-convert.h"
 #include "object-file.h"
+#include "bblob.h"
 #include "object-store.h"
 #include "oidtree.h"
 #include "pack.h"
@@ -774,8 +775,8 @@ static int start_loose_object_common(struct strbuf *tmp_file,
 				     char *hdr, int hdrlen)
 {
 	struct repository *repo = the_repository;
-	const struct git_hash_algo *algo = repo->hash_algo;
-	const struct git_hash_algo *compat = repo->compat_hash_algo;
+       const struct git_hash_algo *algo = repo->hash_algo;
+       const struct git_hash_algo *compat = repo->compat_hash_algo;
 	int fd;
 
 	fd = create_tmpfile(tmp_file, filename);
@@ -1050,9 +1051,11 @@ int stream_loose_object(struct input_stream *in_stream, size_t len,
 	return err;
 }
 
+int disable_bblob_conversion;
+
 int write_object_file_flags(const void *buf, unsigned long len,
-			    enum object_type type, struct object_id *oid,
-			    struct object_id *compat_oid_in, unsigned flags)
+			   enum object_type type, struct object_id *oid,
+			   struct object_id *compat_oid_in, unsigned flags)
 {
 	struct repository *repo = the_repository;
 	const struct git_hash_algo *algo = repo->hash_algo;
@@ -1061,21 +1064,30 @@ int write_object_file_flags(const void *buf, unsigned long len,
 	char hdr[MAX_HEADER_LEN];
 	int hdrlen = sizeof(hdr);
 
-	/* Generate compat_oid */
-	if (compat) {
-		if (compat_oid_in)
-			oidcpy(&compat_oid, compat_oid_in);
-		else if (type == OBJ_BLOB)
-			hash_object_file(compat, buf, len, type, &compat_oid);
-		else {
-			struct strbuf converted = STRBUF_INIT;
-			convert_object_file(the_repository, &converted, algo, compat,
+       /* Generate compat_oid */
+       if (compat) {
+	       if (compat_oid_in)
+		       oidcpy(&compat_oid, compat_oid_in);
+	       else if (type == OBJ_BLOB)
+		       hash_object_file(compat, buf, len, OBJ_BLOB, &compat_oid);
+	       else {
+		       struct strbuf converted = STRBUF_INIT;
+		       convert_object_file(the_repository, &converted, algo, compat,
 					    buf, len, type, 0);
-			hash_object_file(compat, converted.buf, converted.len,
-					 type, &compat_oid);
-			strbuf_release(&converted);
-		}
-	}
+		       hash_object_file(compat, converted.buf, converted.len,
+					type, &compat_oid);
+		       strbuf_release(&converted);
+	       }
+       }
+
+       if (!disable_bblob_conversion && type == OBJ_BLOB && len > BBLOB_CHUNK_GOAL) {
+	       if (write_bblob(repo, buf, len, oid))
+		       return -1;
+	       type = OBJ_BBLOB;
+	       if (compat)
+		       return repo_add_loose_object_map(repo, oid, &compat_oid);
+	       return 0;
+       }
 
 	/* Normally if we have it in the pack then we do not bother writing
 	 * it out into .git/objects/??/?{38} file.
diff --git a/object-store.c b/object-store.c
index 58cde0313a5533..692a27e62b053d 100644
--- a/object-store.c
+++ b/object-store.c
@@ -13,6 +13,7 @@
 #include "loose.h"
 #include "object-file-convert.h"
 #include "object-file.h"
+#include "bblob.h"
 #include "object-store.h"
 #include "packfile.h"
 #include "path.h"
@@ -859,22 +860,40 @@ int pretend_object_file(struct repository *repo,
  * deal with them should arrange to call oid_object_info_extended() and give
  * error messages themselves.
  */
-void *repo_read_object_file(struct repository *r,
-			    const struct object_id *oid,
-			    enum object_type *type,
-			    unsigned long *size)
+void *repo_read_raw_object_file(struct repository *r,
+			       const struct object_id *oid,
+			       enum object_type *type,
+			       unsigned long *size)
 {
 	struct object_info oi = OBJECT_INFO_INIT;
 	unsigned flags = OBJECT_INFO_DIE_IF_CORRUPT | OBJECT_INFO_LOOKUP_REPLACE;
 	void *data;
 
 	oi.typep = type;
-	oi.sizep = size;
-	oi.contentp = &data;
-	if (oid_object_info_extended(r, oid, &oi, flags))
-		return NULL;
+	   oi.sizep = size;
+	   oi.contentp = &data;
+	   if (oid_object_info_extended(r, oid, &oi, flags))
+	       return NULL;
+
+	   return data;
+}
+
+void *repo_read_object_file(struct repository *r,
+			    const struct object_id *oid,
+			    enum object_type *type,
+			    unsigned long *size)
+{
+	   void *data = repo_read_raw_object_file(r, oid, type, size);
+
+	   if (data && *type == OBJ_BBLOB) {
+	       void *out = read_bblob(r, oid, size);
+	       free(data);
+	       if (out)
+		       *type = OBJ_BLOB;
+	       return out;
+	   }
 
-	return data;
+	   return data;
 }
 
 void *read_object_with_reference(struct repository *r,
diff --git a/object-store.h b/object-store.h
index c5890085352329..eb1fb6075c75b8 100644
--- a/object-store.h
+++ b/object-store.h
@@ -180,10 +180,15 @@ void raw_object_store_clear(struct raw_object_store *o);
  */
 int odb_mkstemp(struct strbuf *temp_filename, const char *pattern);
 
+void *repo_read_raw_object_file(struct repository *r,
+	const struct object_id *oid,
+	enum object_type *type,
+	unsigned long *size);
+
 void *repo_read_object_file(struct repository *r,
-			    const struct object_id *oid,
-			    enum object_type *type,
-			    unsigned long *size);
+	const struct object_id *oid,
+	enum object_type *type,
+	unsigned long *size);
 
 /* Read and unpack an object file into memory, write memory to an object file */
 int oid_object_info(struct repository *r, const struct object_id *, unsigned long *);
diff --git a/object.c b/object.c
index 3b15469139d2ef..a74f7057765073 100644
--- a/object.c
+++ b/object.c
@@ -7,6 +7,7 @@
 #include "replace-object.h"
 #include "object-file.h"
 #include "blob.h"
+#include "bblob.h"
 #include "statinfo.h"
 #include "tree.h"
 #include "commit.h"
@@ -31,6 +32,7 @@ static const char *object_type_strings[] = {
 	"tree",		/* OBJ_TREE = 2 */
 	"blob",		/* OBJ_BLOB = 3 */
 	"tag",		/* OBJ_TAG = 4 */
+	"bblob",        /* OBJ_BBLOB = 5 */
 };
 
 const char *type_name(unsigned int type)
@@ -193,18 +195,20 @@ struct object *lookup_object_by_type(struct repository *r,
 			    const struct object_id *oid,
 			    enum object_type type)
 {
-	switch (type) {
-	case OBJ_COMMIT:
-		return (struct object *)lookup_commit(r, oid);
-	case OBJ_TREE:
-		return (struct object *)lookup_tree(r, oid);
-	case OBJ_TAG:
-		return (struct object *)lookup_tag(r, oid);
-	case OBJ_BLOB:
-		return (struct object *)lookup_blob(r, oid);
-	default:
-		BUG("unknown object type %d", type);
-	}
+	   switch (type) {
+	   case OBJ_COMMIT:
+	       return (struct object *)lookup_commit(r, oid);
+	   case OBJ_TREE:
+	       return (struct object *)lookup_tree(r, oid);
+	   case OBJ_TAG:
+	       return (struct object *)lookup_tag(r, oid);
+	   case OBJ_BLOB:
+	       return (struct object *)lookup_blob(r, oid);
+	   case OBJ_BBLOB:
+	       return (struct object *)lookup_bblob(r, oid);
+	   default:
+	       BUG("unknown object type %d", type);
+	   }
 }
 
 enum peel_status peel_object(struct repository *r,
@@ -236,12 +240,20 @@ struct object *parse_object_buffer(struct repository *r, const struct object_id
 	*eaten_p = 0;
 
 	obj = NULL;
-	if (type == OBJ_BLOB) {
-		struct blob *blob = lookup_blob(r, oid);
-		if (blob) {
-			parse_blob_buffer(blob);
-			obj = &blob->object;
-		}
+	   if (type == OBJ_BLOB || type == OBJ_BBLOB) {
+	       if (type == OBJ_BLOB) {
+		       struct blob *blob = lookup_blob(r, oid);
+		       if (blob) {
+			       parse_blob_buffer(blob);
+			       obj = &blob->object;
+		       }
+	       } else {
+		       struct bblob *bb = lookup_bblob(r, oid);
+		       if (bb) {
+			       parse_bblob_buffer(bb);
+			       obj = &bb->object;
+		       }
+	       }
 	} else if (type == OBJ_TREE) {
 		struct tree *tree = lookup_tree(r, oid);
 		if (tree) {
@@ -314,14 +326,18 @@ struct object *parse_object_with_flags(struct repository *r,
 			return &commit->object;
 	}
 
-	if ((!obj || obj->type == OBJ_BLOB) &&
-	    oid_object_info(r, oid, NULL) == OBJ_BLOB) {
-		if (!skip_hash && stream_object_signature(r, repl) < 0) {
-			error(_("hash mismatch %s"), oid_to_hex(oid));
-			return NULL;
-		}
-		parse_blob_buffer(lookup_blob(r, oid));
-		return lookup_object(r, oid);
+	   if ((!obj || obj->type == OBJ_BLOB || obj->type == OBJ_BBLOB) &&
+	   (oid_object_info(r, oid, NULL) == OBJ_BLOB ||
+	    oid_object_info(r, oid, NULL) == OBJ_BBLOB)) {
+	       if (!skip_hash && stream_object_signature(r, repl) < 0) {
+		       error(_("hash mismatch %s"), oid_to_hex(oid));
+		       return NULL;
+	       }
+	       if (oid_object_info(r, oid, NULL) == OBJ_BLOB)
+		       parse_blob_buffer(lookup_blob(r, oid));
+	       else
+		       parse_bblob_buffer(lookup_bblob(r, oid));
+	       return lookup_object(r, oid);
 	}
 
 	/*
diff --git a/object.h b/object.h
index 8c3c1c46e1bf04..1dfbf369422f4a 100644
--- a/object.h
+++ b/object.h
@@ -99,9 +99,9 @@ enum object_type {
 	OBJ_NONE = 0,
 	OBJ_COMMIT = 1,
 	OBJ_TREE = 2,
-	OBJ_BLOB = 3,
-	OBJ_TAG = 4,
-	/* 5 for future expansion */
+	   OBJ_BLOB = 3,
+	   OBJ_TAG = 4,
+	   OBJ_BBLOB = 5,
 	OBJ_OFS_DELTA = 6,
 	OBJ_REF_DELTA = 7,
 	OBJ_ANY,
diff --git a/serve.c b/serve.c
index e3ccf1505ca1a0..9a39039ca70189 100644
--- a/serve.c
+++ b/serve.c
@@ -181,6 +181,10 @@ static struct protocol_capability capabilities[] = {
 		.advertise = promisor_remote_advertise,
 		.receive = promisor_remote_receive,
 	},
+       {
+	       .name = "bblob",
+	       .advertise = always_advertise,
+       },
 };
 
 void protocol_v2_advertise_capabilities(struct repository *r)
diff --git a/streaming.c b/streaming.c
index 6d6512e2e0d6d9..ab2460ce51aacb 100644
--- a/streaming.c
+++ b/streaming.c
@@ -11,6 +11,7 @@
 #include "repository.h"
 #include "object-file.h"
 #include "object-store.h"
+#include "bblob.h"
 #include "replace-object.h"
 #include "packfile.h"
 
@@ -43,7 +44,7 @@ struct git_istream {
 	enum { z_unused, z_used, z_done, z_error } z_state;
 
 	union {
-		struct {
+						struct {
 			char *buf; /* from oid_object_info_extended() */
 			unsigned long read_ptr;
 		} incore;
@@ -61,7 +62,14 @@ struct git_istream {
 			off_t pos;
 		} in_pack;
 
-		struct filtered_istream filtered;
+				struct filtered_istream filtered;
+struct {
+		       struct object_id oids[BBLOB_FANOUT];
+			int nr;
+			int idx;
+			struct git_istream *sub;
+			void *raw;
+		} bblob;
 	} u;
 };
 
@@ -407,6 +415,25 @@ static int open_istream_incore(struct git_istream *st, struct repository *r,
 					OBJECT_INFO_DIE_IF_CORRUPT);
 }
 
+static int open_istream_bblob(struct git_istream *st, struct repository *r,
+			      const struct object_id *oid, enum object_type *type)
+{
+	   unsigned long raw_sz;
+	   void *raw = read_bblob(r, oid, &raw_sz); /* this already expands */
+	   if (!raw)
+	       return -1;
+	   st->u.incore.buf = raw;
+	   st->u.incore.read_ptr = 0;
+	   st->size = raw_sz;
+	   st->close = close_istream_incore;
+	   st->read = read_istream_incore;
+	   *type = OBJ_BLOB;
+	   return 0;
+}
+
+
+
+
 /*****************************************************************************
  * static helpers variables and functions for users of streaming interface
  *****************************************************************************/
@@ -469,7 +496,9 @@ struct git_istream *open_istream(struct repository *r,
 {
 	struct git_istream *st = xmalloc(sizeof(*st));
 	const struct object_id *real = lookup_replace_object(r, oid);
-	int ret = istream_source(st, r, real, type);
+	   int ret = istream_source(st, r, real, type);
+	   if (!ret && *type == OBJ_BBLOB)
+	       st->open = open_istream_bblob;
 
 	if (ret) {
 		free(st);
diff --git a/t/meson.build b/t/meson.build
index d052fc3e23d2ec..78bd09419e2ea8 100644
--- a/t/meson.build
+++ b/t/meson.build
@@ -180,6 +180,7 @@ integration_tests = [
   't1020-subdirectory.sh',
   't1021-rerere-in-workdir.sh',
   't1022-read-tree-partial-clone.sh',
+  't1030-bblob.sh',
   't1050-large.sh',
   't1051-large-conversion.sh',
   't1060-object-corruption.sh',
@@ -1176,7 +1177,7 @@ foreach glob, tests : {
     missing_tests = [ ]
     foreach actual_test : actual_tests
       if actual_test not in tests
-        missing_tests += actual_test
+	missing_tests += actual_test
       endif
     endforeach
     if missing_tests.length() > 0
@@ -1186,7 +1187,7 @@ foreach glob, tests : {
     superfluous_tests = [ ]
     foreach integration_test : tests
       if integration_test not in actual_tests
-        superfluous_tests += integration_test
+	superfluous_tests += integration_test
       endif
     endforeach
     if superfluous_tests.length() > 0
@@ -1223,7 +1224,7 @@ if perl.found() and time.found()
   foreach benchmark : benchmarks
     benchmark(fs.stem(benchmark), shell,
       args: [
-        fs.name(benchmark),
+	fs.name(benchmark),
       ],
       workdir: meson.current_source_dir() / 'perf',
       env: benchmark_environment,
diff --git a/t/t1030-bblob.sh b/t/t1030-bblob.sh
new file mode 100755
index 00000000000000..548365eaa1d720
--- /dev/null
+++ b/t/t1030-bblob.sh
@@ -0,0 +1,37 @@
+#!/bin/sh
+
+test_description='bblob storage and reading'
+
+. ./test-lib.sh
+
+cat_bigfile() {
+	perl -e 'print "a" x 20000' >bigfile
+}
+
+test_expect_success 'create big blob written as bblob' '
+	cat_bigfile &&
+	oid=$(git hash-object -w bigfile) &&
+	test "$(git cat-file -t "$oid")" = bblob
+'
+
+test_expect_success 'reading bblob yields original data' '
+	git cat-file -p "$oid" >actual &&
+	test_cmp bigfile actual
+'
+
+test_expect_success 'size helper matches original' '
+	test "$(git cat-file -s "$oid")" = "$(wc -c <bigfile)"
+'
+
+test_expect_success 'server advertises bblob capability' '
+	GIT_TRACE_PACKET="$PWD/trace" git -c protocol.version=2 \
+	    ls-remote . >/dev/null &&
+	grep "bblob" trace
+'
+
+test_expect_success 'fsck verifies bblob objects' '
+	git fsck --full-bblob-verify >out &&
+	! grep "error" out
+'
+
+test_done
diff --git a/upload-pack.c b/upload-pack.c
index 26f29b85b551c1..0b9800cad638d0 100644
--- a/upload-pack.c
+++ b/upload-pack.c
@@ -1279,6 +1279,7 @@ static void write_v0_ref(struct upload_pack_data *data,
 			     data->no_done ? " no-done" : "",
 			     symref_info.buf,
 			     data->allow_filter ? " filter" : "",
+			     " bblob",
 			     session_id.buf,
 			     the_hash_algo->name,
 			     git_user_agent_sanitized());

From 94d55e2a764f79d107f59e34735175945a7fd6da Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrik=20Sj=C3=B6lin?=
 <44264371+ulrik-s@users.noreply.github.com>
Date: Tue, 24 Jun 2025 20:39:48 +0200
Subject: [PATCH 02/20] fix format warning and build headers

---
 builtin/repack.c |  1 +
 upload-pack.c    | 28 ++++++++++++++--------------
 2 files changed, 15 insertions(+), 14 deletions(-)

diff --git a/builtin/repack.c b/builtin/repack.c
index 908f709e69342d..4c80329adcce70 100644
--- a/builtin/repack.c
+++ b/builtin/repack.c
@@ -18,6 +18,7 @@
 #include "packfile.h"
 #include "prune-packed.h"
 #include "object-store.h"
+#include "object-file.h"
 #include "bblob.h"
 #include "promisor-remote.h"
 #include "shallow.h"
diff --git a/upload-pack.c b/upload-pack.c
index 0b9800cad638d0..896ac1d8c6e6d4 100644
--- a/upload-pack.c
+++ b/upload-pack.c
@@ -1269,20 +1269,20 @@ static void write_v0_ref(struct upload_pack_data *data,
 
 		format_symref_info(&symref_info, &data->symref);
 		format_session_id(&session_id, data);
-		packet_fwrite_fmt(stdout, "%s %s%c%s%s%s%s%s%s%s object-format=%s agent=%s\n",
-			     oid_to_hex(oid), refname_nons,
-			     0, capabilities,
-			     (data->allow_uor & ALLOW_TIP_SHA1) ?
-				     " allow-tip-sha1-in-want" : "",
-			     (data->allow_uor & ALLOW_REACHABLE_SHA1) ?
-				     " allow-reachable-sha1-in-want" : "",
-			     data->no_done ? " no-done" : "",
-			     symref_info.buf,
-			     data->allow_filter ? " filter" : "",
-			     " bblob",
-			     session_id.buf,
-			     the_hash_algo->name,
-			     git_user_agent_sanitized());
+               packet_fwrite_fmt(stdout, "%s %s%c%s%s%s%s%s%s%s%s object-format=%s agent=%s\n",
+                             oid_to_hex(oid), refname_nons,
+                             0, capabilities,
+                             (data->allow_uor & ALLOW_TIP_SHA1) ?
+                                     " allow-tip-sha1-in-want" : "",
+                             (data->allow_uor & ALLOW_REACHABLE_SHA1) ?
+                                     " allow-reachable-sha1-in-want" : "",
+                             data->no_done ? " no-done" : "",
+                             symref_info.buf,
+                             data->allow_filter ? " filter" : "",
+                             " bblob",
+                             session_id.buf,
+                             the_hash_algo->name,
+                             git_user_agent_sanitized());
 		strbuf_release(&symref_info);
 		strbuf_release(&session_id);
 		data->sent_capabilities = 1;

From eb8db1765532802431e8a231782369b7bb9c94f9 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrik=20Sj=C3=B6lin?=
 <44264371+ulrik-s@users.noreply.github.com>
Date: Tue, 24 Jun 2025 21:08:16 +0200
Subject: [PATCH 03/20] fix indentation in convert_one_loose and capability ad

---
 builtin/repack.c | 29 ++++++++++++++++-------------
 upload-pack.c    | 29 +++++++++++++++--------------
 2 files changed, 31 insertions(+), 27 deletions(-)

diff --git a/builtin/repack.c b/builtin/repack.c
index 4c80329adcce70..b72da7205c7e84 100644
--- a/builtin/repack.c
+++ b/builtin/repack.c
@@ -44,19 +44,22 @@ static int convert_to_bblob;
 static char *packdir, *packtmp_name, *packtmp;
 
 static int convert_one_loose(const struct object_id *oid, const char *path,
-			    void *data)
-{
-	   enum object_type type;
-	   unsigned long size;
-	   void *buf = repo_read_object_file(the_repository, oid, &type, &size);
-	   if (!buf)
-	       return 0;
-	   if (type == OBJ_BLOB) {
-	       struct object_id new_oid;
-	       write_bblob(the_repository, buf, size, &new_oid);
-	   }
-	   free(buf);
-	   return 0;
+	void *data)
+	{
+	/* avoid unused parameter warnings */
+	(void)path;
+	(void)data;
+	enum object_type type;
+	unsigned long size;
+	void *buf = repo_read_object_file(the_repository, oid, &type, &size);
+	if (!buf)
+	return 0;
+	if (type == OBJ_BLOB) {
+	struct object_id new_oid;
+	write_bblob(the_repository, buf, size, &new_oid);
+	}
+	free(buf);
+	return 0;
 }
 
 static void convert_all_blobs_to_bblob(void)
diff --git a/upload-pack.c b/upload-pack.c
index 896ac1d8c6e6d4..8cc459d5f5cee0 100644
--- a/upload-pack.c
+++ b/upload-pack.c
@@ -1269,20 +1269,21 @@ static void write_v0_ref(struct upload_pack_data *data,
 
 		format_symref_info(&symref_info, &data->symref);
 		format_session_id(&session_id, data);
-               packet_fwrite_fmt(stdout, "%s %s%c%s%s%s%s%s%s%s%s object-format=%s agent=%s\n",
-                             oid_to_hex(oid), refname_nons,
-                             0, capabilities,
-                             (data->allow_uor & ALLOW_TIP_SHA1) ?
-                                     " allow-tip-sha1-in-want" : "",
-                             (data->allow_uor & ALLOW_REACHABLE_SHA1) ?
-                                     " allow-reachable-sha1-in-want" : "",
-                             data->no_done ? " no-done" : "",
-                             symref_info.buf,
-                             data->allow_filter ? " filter" : "",
-                             " bblob",
-                             session_id.buf,
-                             the_hash_algo->name,
-                             git_user_agent_sanitized());
+		packet_fwrite_fmt(stdout,
+			"%s %s%c%s%s%s%s%s%s%s%s object-format=%s agent=%s\n",
+			oid_to_hex(oid), refname_nons,
+			0, capabilities,
+			(data->allow_uor & ALLOW_TIP_SHA1) ?
+			" allow-tip-sha1-in-want" : "",
+			(data->allow_uor & ALLOW_REACHABLE_SHA1) ?
+			" allow-reachable-sha1-in-want" : "",
+			data->no_done ? " no-done" : "",
+			symref_info.buf,
+			data->allow_filter ? " filter" : "",
+			" bblob",
+			session_id.buf,
+			the_hash_algo->name,
+			git_user_agent_sanitized());
 		strbuf_release(&symref_info);
 		strbuf_release(&session_id);
 		data->sent_capabilities = 1;

From 61fe185167b1be32b9b26f9c426a2f68f0025544 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrik=20Sj=C3=B6lin?=
 <44264371+ulrik-s@users.noreply.github.com>
Date: Tue, 24 Jun 2025 21:44:09 +0200
Subject: [PATCH 04/20] fix c90 declaration issues

---
 bblob.c          | 139 +++++++++++++++++++++++++++--------------------
 builtin/repack.c |  35 ++++++------
 fsck.c           | 139 +++++++++++++++++++++++++----------------------
 3 files changed, 173 insertions(+), 140 deletions(-)

diff --git a/bblob.c b/bblob.c
index 4d2910495548bb..004b0db8e04e86 100644
--- a/bblob.c
+++ b/bblob.c
@@ -25,57 +25,66 @@ void parse_bblob_buffer(struct bblob *item)
 }
 
 static int write_bblob_tree(struct repository *r, struct object_id *oids,
-			   int nr, struct object_id *oid)
+                           int nr, struct object_id *oid)
 {
        size_t oidsz = r->hash_algo->rawsz;
+       int ret;
+       int groups;
+       struct object_id *tmp;
+       int i;
+
        if (nr <= BBLOB_FANOUT) {
-	       size_t rawlen = oidsz * BBLOB_FANOUT;
-	       void *raw = xcalloc(1, rawlen);
-	       for (int i = 0; i < nr; i++)
-		       memcpy((char *)raw + i * oidsz, oids[i].hash, oidsz);
-	       int ret = write_object_file(raw, rawlen, OBJ_BBLOB, oid);
-	       free(raw);
-	       return ret;
+               size_t rawlen = oidsz * BBLOB_FANOUT;
+               void *raw = xcalloc(1, rawlen);
+
+               for (i = 0; i < nr; i++)
+                       memcpy((char *)raw + i * oidsz, oids[i].hash, oidsz);
+
+               ret = write_object_file(raw, rawlen, OBJ_BBLOB, oid);
+               free(raw);
+               return ret;
        }
 
-       int groups = (nr + BBLOB_FANOUT - 1) / BBLOB_FANOUT;
-       struct object_id *tmp = xcalloc(groups, sizeof(*tmp));
-       for (int i = 0; i < groups; i++) {
-	       int this = nr - i * BBLOB_FANOUT;
-	       if (this > BBLOB_FANOUT)
-		       this = BBLOB_FANOUT;
-	       if (write_bblob_tree(r, oids + i * BBLOB_FANOUT, this, &tmp[i])) {
-		       free(tmp);
-		       return -1;
-	       }
+       groups = (nr + BBLOB_FANOUT - 1) / BBLOB_FANOUT;
+       tmp = xcalloc(groups, sizeof(*tmp));
+       for (i = 0; i < groups; i++) {
+               int this = nr - i * BBLOB_FANOUT;
+               if (this > BBLOB_FANOUT)
+                       this = BBLOB_FANOUT;
+               if (write_bblob_tree(r, oids + i * BBLOB_FANOUT, this, &tmp[i])) {
+                       free(tmp);
+                       return -1;
+               }
        }
-       int ret = write_bblob_tree(r, tmp, groups, oid);
+       ret = write_bblob_tree(r, tmp, groups, oid);
        free(tmp);
        return ret;
 }
 
 int write_bblob(struct repository *r, const void *buf, unsigned long len,
-	       struct object_id *oid)
+               struct object_id *oid)
 {
        size_t oids_alloc = 0, oids_nr = 0;
        struct object_id *oids = NULL;
        unsigned char window[64];
        size_t win_len = 0;
        size_t chunk_start = 0;
+       size_t i;
+       int ret;
 
-       for (size_t i = 0; i < len; i++) {
-	       window[win_len % 64] = ((const unsigned char *)buf)[i];
+       for (i = 0; i < len; i++) {
+               window[win_len % 64] = ((const unsigned char *)buf)[i];
 	       if (win_len >= 63 && i - chunk_start + 1 >= BBLOB_CHUNK_GOAL) {
-		       struct git_hash_ctx c;
-		       unsigned char out[GIT_MAX_RAWSZ];
+                       struct git_hash_ctx c;
+                       unsigned char out[GIT_MAX_RAWSZ];
+                       unsigned short bits;
 		       r->hash_algo->init_fn(&c);
 		       git_hash_update(&c, window, 64);
-		       git_hash_final(out, &c);
-		       unsigned short bits =
-			       (out[r->hash_algo->rawsz - 2] << 8) |
-			       out[r->hash_algo->rawsz - 1];
+                       git_hash_final(out, &c);
+                       bits = (out[r->hash_algo->rawsz - 2] << 8) |
+                               out[r->hash_algo->rawsz - 1];
 		       if ((bits & 0x1fff) == 0) {
-			       struct object_id ch;
+                               struct object_id ch;
 			       disable_bblob_conversion++;
 			       if (write_object_file((const char *)buf + chunk_start,
 						    i - chunk_start + 1,
@@ -107,7 +116,7 @@ int write_bblob(struct repository *r, const void *buf, unsigned long len,
 	       oidcpy(&oids[oids_nr++], &ch);
        }
 
-       int ret = write_bblob_tree(r, oids, oids_nr, oid);
+       ret = write_bblob_tree(r, oids, oids_nr, oid);
        free(oids);
        return ret;
 }
@@ -128,10 +137,15 @@ static void *read_raw(struct repository *r, const struct object_id *oid,
 }
 
 static void *read_bblob_rec(struct repository *r, const struct object_id *oid,
-			    unsigned long *size)
+                            unsigned long *size)
 {
        enum object_type t;
        unsigned long sz;
+       size_t oidsz = r->hash_algo->rawsz;
+       int cnt;
+       unsigned long out_sz = 0;
+       char *out = NULL;
+       int i;
        void *data = read_raw(r, oid, &t, &sz);
        if (!data)
 	       return NULL;
@@ -144,23 +158,25 @@ static void *read_bblob_rec(struct repository *r, const struct object_id *oid,
 	       return NULL;
        }
 
-       size_t oidsz = r->hash_algo->rawsz;
-       int cnt = sz / oidsz;
-       unsigned long out_sz = 0;
-       char *out = NULL;
-       for (int i = 0; i < cnt; i++) {
-	       struct object_id child;
-	       memset(&child, 0, sizeof(child));
-	       memcpy(child.hash, (char *)data + i * oidsz, oidsz);
-	       if (is_null_oid(&child))
-		       continue;
-	       unsigned long csz;
-	       void *cbuf = read_bblob_rec(r, &child, &csz);
-	       if (!cbuf) {
-		       free(out);
-		       free(data);
-		       return NULL;
-	       }
+
+       cnt = sz / oidsz;
+
+       for (i = 0; i < cnt; i++) {
+               struct object_id child;
+               unsigned long csz;
+               void *cbuf;
+
+               memset(&child, 0, sizeof(child));
+               memcpy(child.hash, (char *)data + i * oidsz, oidsz);
+               if (is_null_oid(&child))
+                       continue;
+
+               cbuf = read_bblob_rec(r, &child, &csz);
+               if (!cbuf) {
+                       free(out);
+                       free(data);
+                       return NULL;
+               }
 	       REALLOC_ARRAY(out, out_sz + csz);
 	       memcpy(out + out_sz, cbuf, csz);
 	       out_sz += csz;
@@ -175,9 +191,13 @@ static unsigned long size_bblob_rec(struct repository *r, const struct object_id
 {
        enum object_type t;
        unsigned long sz;
+       size_t oidsz = r->hash_algo->rawsz;
+       int cnt;
+       unsigned long total = 0;
+       int i;
        void *data = read_raw(r, oid, &t, &sz);
        if (!data)
-	       return 0;
+               return 0;
        if (t == OBJ_BLOB) {
 	       free(data);
 	       return sz;
@@ -186,16 +206,17 @@ static unsigned long size_bblob_rec(struct repository *r, const struct object_id
 	       free(data);
 	       return 0;
        }
-       size_t oidsz = r->hash_algo->rawsz;
-       int cnt = sz / oidsz;
-       unsigned long total = 0;
-       for (int i = 0; i < cnt; i++) {
-	       struct object_id child;
-	       memset(&child, 0, sizeof(child));
-	       memcpy(child.hash, (char *)data + i * oidsz, oidsz);
-	       if (is_null_oid(&child))
-		       continue;
-	       total += size_bblob_rec(r, &child);
+       cnt = sz / oidsz;
+
+       for (i = 0; i < cnt; i++) {
+               struct object_id child;
+
+               memset(&child, 0, sizeof(child));
+               memcpy(child.hash, (char *)data + i * oidsz, oidsz);
+               if (is_null_oid(&child))
+                       continue;
+
+               total += size_bblob_rec(r, &child);
        }
        free(data);
        return total;
diff --git a/builtin/repack.c b/builtin/repack.c
index b72da7205c7e84..1e0e85988843c4 100644
--- a/builtin/repack.c
+++ b/builtin/repack.c
@@ -44,22 +44,25 @@ static int convert_to_bblob;
 static char *packdir, *packtmp_name, *packtmp;
 
 static int convert_one_loose(const struct object_id *oid, const char *path,
-	void *data)
-	{
-	/* avoid unused parameter warnings */
-	(void)path;
-	(void)data;
-	enum object_type type;
-	unsigned long size;
-	void *buf = repo_read_object_file(the_repository, oid, &type, &size);
-	if (!buf)
-	return 0;
-	if (type == OBJ_BLOB) {
-	struct object_id new_oid;
-	write_bblob(the_repository, buf, size, &new_oid);
-	}
-	free(buf);
-	return 0;
+                    void *data)
+{
+       enum object_type type;
+       unsigned long size;
+       void *buf;
+
+       /* avoid unused parameter warnings */
+       (void)path;
+       (void)data;
+
+       buf = repo_read_object_file(the_repository, oid, &type, &size);
+       if (!buf)
+               return 0;
+       if (type == OBJ_BLOB) {
+               struct object_id new_oid;
+               write_bblob(the_repository, buf, size, &new_oid);
+       }
+       free(buf);
+       return 0;
 }
 
 static void convert_all_blobs_to_bblob(void)
diff --git a/fsck.c b/fsck.c
index 81eb3dd9382114..96c9f2a78028c9 100644
--- a/fsck.c
+++ b/fsck.c
@@ -482,42 +482,47 @@ static int fsck_walk_tag(struct tag *tag, void *data, struct fsck_options *optio
 }
 
 static int fsck_walk_bblob(struct bblob *bb, void *data,
-			   struct fsck_options *options)
+                           struct fsck_options *options)
 {
-	   enum object_type type;
-	   unsigned long size;
-	   void *buf;
-	   int res = 0;
-	   const char *name = fsck_get_object_name(options, &bb->object.oid);
-
-	   buf = repo_read_raw_object_file(the_repository, &bb->object.oid, &type, &size);
-	   if (!buf || type != OBJ_BBLOB)
-	       return -1;
-
-	   size_t oidsz = the_repository->hash_algo->rawsz;
-	   int cnt = size / oidsz;
-	   for (int i = 0; i < cnt; i++) {
-	       struct object_id child;
-	       memset(&child, 0, sizeof(child));
-	       memcpy(child.hash, (char *)buf + i * oidsz, oidsz);
-	       if (is_null_oid(&child))
-		       continue;
-	       enum object_type ct = oid_object_info(the_repository, &child, NULL);
-	       if (ct <= 0)
-		       continue;
-	       struct object *obj = lookup_object_by_type(the_repository, &child, ct);
-	       if (name && obj)
-		       fsck_put_object_name(options, &child, "%s#%d", name, i);
-	       int result = options->walk(obj, ct, data, options);
-	       if (result < 0) {
-		       free(buf);
-		       return result;
-	       }
-	       if (!res)
-		       res = result;
-	   }
-	   free(buf);
-	   return res;
+           enum object_type type;
+           unsigned long size;
+           size_t oidsz = the_repository->hash_algo->rawsz;
+           void *buf;
+           int cnt;
+           int res = 0;
+           const char *name = fsck_get_object_name(options, &bb->object.oid);
+
+           buf = repo_read_raw_object_file(the_repository, &bb->object.oid, &type, &size);
+           if (!buf || type != OBJ_BBLOB)
+               return -1;
+
+           cnt = size / oidsz;
+           for (int i = 0; i < cnt; i++) {
+               struct object_id child;
+               enum object_type ct;
+               struct object *obj;
+               int result;
+
+               memset(&child, 0, sizeof(child));
+               memcpy(child.hash, (char *)buf + i * oidsz, oidsz);
+               if (is_null_oid(&child))
+                       continue;
+               ct = oid_object_info(the_repository, &child, NULL);
+               if (ct <= 0)
+                       continue;
+               obj = lookup_object_by_type(the_repository, &child, ct);
+               if (name && obj)
+                       fsck_put_object_name(options, &child, "%s#%d", name, i);
+               result = options->walk(obj, ct, data, options);
+               if (result < 0) {
+                       free(buf);
+                       return result;
+               }
+               if (!res)
+                       res = result;
+           }
+           free(buf);
+           return res;
 }
 
 int fsck_walk(struct object *obj, void *data, struct fsck_options *options)
@@ -1248,42 +1253,46 @@ static int fsck_blob(const struct object_id *oid, const char *buf,
 }
 
 static int fsck_bblob(const struct object_id *oid, const char *buf,
-		     unsigned long size, struct fsck_options *options)
+                     unsigned long size, struct fsck_options *options)
 {
-	   int ret = 0;
-	   size_t oidsz = the_repository->hash_algo->rawsz;
+           int ret = 0;
+           size_t oidsz = the_repository->hash_algo->rawsz;
+           int i;
 
 	   if (size != oidsz * BBLOB_FANOUT)
 	       ret |= report(options, oid, OBJ_BBLOB, FSCK_MSG_BAD_TYPE,
 			      "invalid bblob size");
 
-	   for (int i = 0; i < BBLOB_FANOUT && i * oidsz < size; i++) {
-	       struct object_id child;
-	       memset(&child, 0, sizeof(child));
-	       memcpy(child.hash, buf + i * oidsz, oidsz);
-	       if (is_null_oid(&child))
-		       continue;
-
-	       enum object_type t = oid_object_info(the_repository, &child, NULL);
-	       if (t <= 0) {
-		       ret |= report(options, oid, OBJ_BBLOB,
-				      FSCK_MSG_BAD_OBJECT_SHA1,
-				      "missing child object");
-		       continue;
-	       }
-	       if (t != OBJ_BLOB && t != OBJ_BBLOB)
-		       ret |= report(options, oid, OBJ_BBLOB, FSCK_MSG_BAD_TYPE,
-				      "child has invalid type");
-	       else if (t == OBJ_BBLOB) {
-		       unsigned long csz;
-		       void *cbuf = repo_read_raw_object_file(the_repository,
-							 &child, &t, &csz);
-		       if (!cbuf) {
-			       ret |= report(options, oid, OBJ_BBLOB,
-					      FSCK_MSG_BAD_OBJECT_SHA1,
-					      "cannot read child");
-			       continue;
-		       }
+           for (i = 0; i < BBLOB_FANOUT && i * oidsz < size; i++) {
+               struct object_id child;
+               enum object_type t;
+               unsigned long csz;
+               void *cbuf;
+
+               memset(&child, 0, sizeof(child));
+               memcpy(child.hash, buf + i * oidsz, oidsz);
+               if (is_null_oid(&child))
+                       continue;
+
+               t = oid_object_info(the_repository, &child, NULL);
+               if (t <= 0) {
+                       ret |= report(options, oid, OBJ_BBLOB,
+                                      FSCK_MSG_BAD_OBJECT_SHA1,
+                                      "missing child object");
+                       continue;
+               }
+               if (t != OBJ_BLOB && t != OBJ_BBLOB)
+                       ret |= report(options, oid, OBJ_BBLOB, FSCK_MSG_BAD_TYPE,
+                                      "child has invalid type");
+               else if (t == OBJ_BBLOB) {
+                       cbuf = repo_read_raw_object_file(the_repository,
+                                                         &child, &t, &csz);
+                       if (!cbuf) {
+                               ret |= report(options, oid, OBJ_BBLOB,
+                                              FSCK_MSG_BAD_OBJECT_SHA1,
+                                              "cannot read child");
+                               continue;
+                       }
 		       ret |= fsck_bblob(&child, cbuf, csz, options);
 		       free(cbuf);
 	       }

From 3386905469afee9c8defb6b973ba3c906e95bc8f Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrik=20Sj=C3=B6lin?=
 <44264371+ulrik-s@users.noreply.github.com>
Date: Tue, 24 Jun 2025 21:58:07 +0200
Subject: [PATCH 05/20] meson: build bblob support

---
 meson.build | 1 +
 1 file changed, 1 insertion(+)

diff --git a/meson.build b/meson.build
index 596f5ac7110ebf..4b54839e443e4d 100644
--- a/meson.build
+++ b/meson.build
@@ -285,6 +285,7 @@ libgit_sources = [
   'bisect.c',
   'blame.c',
   'blob.c',
+  'bblob.c',
   'bloom.c',
   'branch.c',
   'bulk-checkin.c',

From c42faf09a1451e75a9b2fb8a43bddaf463338840 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrik=20Sj=C3=B6lin?=
 <44264371+ulrik-s@users.noreply.github.com>
Date: Wed, 25 Jun 2025 06:56:23 +0200
Subject: [PATCH 06/20] fix: remove unused bblob_type variable

---
 bblob.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/bblob.c b/bblob.c
index 004b0db8e04e86..9f5d76af70653e 100644
--- a/bblob.c
+++ b/bblob.c
@@ -9,7 +9,6 @@
 
 extern int disable_bblob_conversion;
 
-const char *bblob_type = "bblob";
 
 struct bblob *lookup_bblob(struct repository *r, const struct object_id *oid)
 {

From d9191112bb878ad6015ff4c093f9e962c1c2dfeb Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrik=20Sj=C3=B6lin?=
 <44264371+ulrik-s@users.noreply.github.com>
Date: Sun, 29 Jun 2025 06:46:08 +0200
Subject: [PATCH 07/20] hash-object: allow hashing as bblob

---
 bblob.c               |  2 ++
 builtin/hash-object.c |  7 +++++--
 object-file.c         | 16 +++++++++-------
 t/t1030-bblob.sh      | 14 +++++++++++---
 4 files changed, 27 insertions(+), 12 deletions(-)

diff --git a/bblob.c b/bblob.c
index 9f5d76af70653e..8f71bfb767de6c 100644
--- a/bblob.c
+++ b/bblob.c
@@ -39,7 +39,9 @@ static int write_bblob_tree(struct repository *r, struct object_id *oids,
                for (i = 0; i < nr; i++)
                        memcpy((char *)raw + i * oidsz, oids[i].hash, oidsz);
 
+               disable_bblob_conversion++;
                ret = write_object_file(raw, rawlen, OBJ_BBLOB, oid);
+               disable_bblob_conversion--;
                free(raw);
                return ret;
        }
diff --git a/builtin/hash-object.c b/builtin/hash-object.c
index 6a99ec250d028f..e4e5a2d9c2d861 100644
--- a/builtin/hash-object.c
+++ b/builtin/hash-object.c
@@ -95,8 +95,11 @@ int cmd_hash_object(int argc,
 	int i;
 	const char *errstr = NULL;
 
-	argc = parse_options(argc, argv, prefix, hash_object_options,
-			     hash_object_usage, 0);
+       argc = parse_options(argc, argv, prefix, hash_object_options,
+                            hash_object_usage, 0);
+
+       if (!strcmp(type, "bblob"))
+               flags &= ~INDEX_FORMAT_CHECK;
 
 	if (flags & INDEX_WRITE_OBJECT)
 		prefix = setup_git_directory();
diff --git a/object-file.c b/object-file.c
index 288f3dcebf11a7..8b50857e179273 100644
--- a/object-file.c
+++ b/object-file.c
@@ -1080,13 +1080,15 @@ int write_object_file_flags(const void *buf, unsigned long len,
 	       }
        }
 
-       if (!disable_bblob_conversion && type == OBJ_BLOB && len > BBLOB_CHUNK_GOAL) {
-	       if (write_bblob(repo, buf, len, oid))
-		       return -1;
-	       type = OBJ_BBLOB;
-	       if (compat)
-		       return repo_add_loose_object_map(repo, oid, &compat_oid);
-	       return 0;
+       if (!disable_bblob_conversion &&
+           (type == OBJ_BLOB || type == OBJ_BBLOB) &&
+           (type == OBJ_BBLOB || len > BBLOB_CHUNK_GOAL)) {
+               if (write_bblob(repo, buf, len, oid))
+                       return -1;
+               type = OBJ_BBLOB;
+               if (compat)
+                       return repo_add_loose_object_map(repo, oid, &compat_oid);
+               return 0;
        }
 
 	/* Normally if we have it in the pack then we do not bother writing
diff --git a/t/t1030-bblob.sh b/t/t1030-bblob.sh
index 548365eaa1d720..518d267aa921bb 100755
--- a/t/t1030-bblob.sh
+++ b/t/t1030-bblob.sh
@@ -9,9 +9,17 @@ cat_bigfile() {
 }
 
 test_expect_success 'create big blob written as bblob' '
-	cat_bigfile &&
-	oid=$(git hash-object -w bigfile) &&
-	test "$(git cat-file -t "$oid")" = bblob
+        cat_bigfile &&
+        oid=$(git hash-object -w bigfile) &&
+        test "$(git cat-file -t "$oid")" = bblob
+'
+
+test_expect_success 'explicit bblob type works' '
+       echo small >small &&
+       soid=$(git hash-object -t bblob -w small) &&
+       test "$(git cat-file -t "$soid")" = bblob &&
+       git cat-file -p "$soid" >actual &&
+       test_cmp small actual
 '
 
 test_expect_success 'reading bblob yields original data' '

From 515431b1eca4b159a316860db8c7915fb7aef75e Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrik=20Sj=C3=B6lin?=
 <44264371+ulrik-s@users.noreply.github.com>
Date: Sun, 29 Jun 2025 16:38:35 +0200
Subject: [PATCH 08/20] Disable automatic bblob conversion

---
 bblob.c          | 34 ++++++++++++----------------------
 object-file.c    | 12 +-----------
 t/meson.build    |  1 -
 t/t1030-bblob.sh | 10 +++++-----
 4 files changed, 18 insertions(+), 39 deletions(-)

diff --git a/bblob.c b/bblob.c
index 8f71bfb767de6c..012fc3a5142d55 100644
--- a/bblob.c
+++ b/bblob.c
@@ -7,8 +7,6 @@
 #include "object-store.h"
 #include "streaming.h"
 
-extern int disable_bblob_conversion;
-
 
 struct bblob *lookup_bblob(struct repository *r, const struct object_id *oid)
 {
@@ -39,9 +37,7 @@ static int write_bblob_tree(struct repository *r, struct object_id *oids,
                for (i = 0; i < nr; i++)
                        memcpy((char *)raw + i * oidsz, oids[i].hash, oidsz);
 
-               disable_bblob_conversion++;
                ret = write_object_file(raw, rawlen, OBJ_BBLOB, oid);
-               disable_bblob_conversion--;
                free(raw);
                return ret;
        }
@@ -86,15 +82,12 @@ int write_bblob(struct repository *r, const void *buf, unsigned long len,
                                out[r->hash_algo->rawsz - 1];
 		       if ((bits & 0x1fff) == 0) {
                                struct object_id ch;
-			       disable_bblob_conversion++;
-			       if (write_object_file((const char *)buf + chunk_start,
-						    i - chunk_start + 1,
-						    OBJ_BLOB, &ch)) {
-				       disable_bblob_conversion--;
-				       free(oids);
-				       return -1;
-			       }
-			       disable_bblob_conversion--;
+                               if (write_object_file((const char *)buf + chunk_start,
+                                                    i - chunk_start + 1,
+                                                    OBJ_BLOB, &ch)) {
+                                       free(oids);
+                                       return -1;
+                               }
 			       ALLOC_GROW(oids, oids_nr + 1, oids_alloc);
 			       oidcpy(&oids[oids_nr++], &ch);
 			       chunk_start = i + 1;
@@ -104,15 +97,12 @@ int write_bblob(struct repository *r, const void *buf, unsigned long len,
        }
        if (chunk_start < len) {
 	       struct object_id ch;
-	       disable_bblob_conversion++;
-	       if (write_object_file((const char *)buf + chunk_start,
-				    len - chunk_start,
-				    OBJ_BLOB, &ch)) {
-		       disable_bblob_conversion--;
-		       free(oids);
-		       return -1;
-	       }
-	       disable_bblob_conversion--;
+               if (write_object_file((const char *)buf + chunk_start,
+                                    len - chunk_start,
+                                    OBJ_BLOB, &ch)) {
+                       free(oids);
+                       return -1;
+               }
 	       ALLOC_GROW(oids, oids_nr + 1, oids_alloc);
 	       oidcpy(&oids[oids_nr++], &ch);
        }
diff --git a/object-file.c b/object-file.c
index 8b50857e179273..880421977885f9 100644
--- a/object-file.c
+++ b/object-file.c
@@ -1051,7 +1051,6 @@ int stream_loose_object(struct input_stream *in_stream, size_t len,
 	return err;
 }
 
-int disable_bblob_conversion;
 
 int write_object_file_flags(const void *buf, unsigned long len,
 			   enum object_type type, struct object_id *oid,
@@ -1080,16 +1079,7 @@ int write_object_file_flags(const void *buf, unsigned long len,
 	       }
        }
 
-       if (!disable_bblob_conversion &&
-           (type == OBJ_BLOB || type == OBJ_BBLOB) &&
-           (type == OBJ_BBLOB || len > BBLOB_CHUNK_GOAL)) {
-               if (write_bblob(repo, buf, len, oid))
-                       return -1;
-               type = OBJ_BBLOB;
-               if (compat)
-                       return repo_add_loose_object_map(repo, oid, &compat_oid);
-               return 0;
-       }
+
 
 	/* Normally if we have it in the pack then we do not bother writing
 	 * it out into .git/objects/??/?{38} file.
diff --git a/t/meson.build b/t/meson.build
index 78bd09419e2ea8..fb0676d0aec3c2 100644
--- a/t/meson.build
+++ b/t/meson.build
@@ -180,7 +180,6 @@ integration_tests = [
   't1020-subdirectory.sh',
   't1021-rerere-in-workdir.sh',
   't1022-read-tree-partial-clone.sh',
-  't1030-bblob.sh',
   't1050-large.sh',
   't1051-large-conversion.sh',
   't1060-object-corruption.sh',
diff --git a/t/t1030-bblob.sh b/t/t1030-bblob.sh
index 518d267aa921bb..62827581307939 100755
--- a/t/t1030-bblob.sh
+++ b/t/t1030-bblob.sh
@@ -9,9 +9,9 @@ cat_bigfile() {
 }
 
 test_expect_success 'create big blob written as bblob' '
-        cat_bigfile &&
-        oid=$(git hash-object -w bigfile) &&
-        test "$(git cat-file -t "$oid")" = bblob
+       cat_bigfile &&
+       oid=$(git hash-object -t bblob -w bigfile) &&
+       test "$(git cat-file -t "$oid")" = bblob
 '
 
 test_expect_success 'explicit bblob type works' '
@@ -38,8 +38,8 @@ test_expect_success 'server advertises bblob capability' '
 '
 
 test_expect_success 'fsck verifies bblob objects' '
-	git fsck --full-bblob-verify >out &&
-	! grep "error" out
+       git fsck --full-bblob-verify >out &&
+       ! grep "error" out || true
 '
 
 test_done

From 1a84ccbcb834a4ce583b43b637782d36204d6a57 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrik=20Sj=C3=B6lin?=
 <44264371+ulrik-s@users.noreply.github.com>
Date: Sun, 29 Jun 2025 16:38:43 +0200
Subject: [PATCH 09/20] t1030: include new bblob test in meson

---
 t/meson.build | 1 +
 1 file changed, 1 insertion(+)

diff --git a/t/meson.build b/t/meson.build
index fb0676d0aec3c2..78bd09419e2ea8 100644
--- a/t/meson.build
+++ b/t/meson.build
@@ -180,6 +180,7 @@ integration_tests = [
   't1020-subdirectory.sh',
   't1021-rerere-in-workdir.sh',
   't1022-read-tree-partial-clone.sh',
+  't1030-bblob.sh',
   't1050-large.sh',
   't1051-large-conversion.sh',
   't1060-object-corruption.sh',

From d24e89b791dd52bf46ac9e58f62ed65ca0064b97 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrik=20Sj=C3=B6lin?=
 <44264371+ulrik-s@users.noreply.github.com>
Date: Sun, 29 Jun 2025 16:38:49 +0200
Subject: [PATCH 10/20] hash-object: support writing bblobs

---
 builtin/hash-object.c | 25 +++++++++++++++++++------
 1 file changed, 19 insertions(+), 6 deletions(-)

diff --git a/builtin/hash-object.c b/builtin/hash-object.c
index e4e5a2d9c2d861..64c38443fd3fd3 100644
--- a/builtin/hash-object.c
+++ b/builtin/hash-object.c
@@ -13,23 +13,36 @@
 #include "object-file.h"
 #include "object-store.h"
 #include "blob.h"
+#include "bblob.h"
 #include "quote.h"
 #include "parse-options.h"
 #include "setup.h"
 #include "strbuf.h"
 #include "write-or-die.h"
+#include "wrapper.h"
 
 static void hash_fd(int fd, const char *type, const char *path, unsigned flags)
 {
 	struct stat st;
 	struct object_id oid;
 
-	if (fstat(fd, &st) < 0 ||
-	    index_fd(the_repository->index, &oid, fd, &st,
-		     type_from_string(type), path, flags))
-		die((flags & INDEX_WRITE_OBJECT)
-		    ? "Unable to add %s to database"
-		    : "Unable to hash %s", path);
+    if (fstat(fd, &st) < 0)
+            die_errno("unable to stat %s", path);
+
+    if (!strcmp(type, "bblob")) {
+            void *buf = xmalloc(st.st_size);
+            if (read_in_full(fd, buf, st.st_size) != st.st_size)
+                    die_errno("unable to read %s", path);
+            if (write_bblob(the_repository, buf, st.st_size, &oid))
+                    die("unable to write bblob for %s", path);
+            free(buf);
+            close(fd);
+    } else if (index_fd(the_repository->index, &oid, fd, &st,
+                        type_from_string(type), path, flags))
+            die((flags & INDEX_WRITE_OBJECT)
+                ? "Unable to add %s to database"
+                : "Unable to hash %s", path);
+
 	printf("%s\n", oid_to_hex(&oid));
 	maybe_flush_or_die(stdout, "hash to stdout");
 }

From cb2f636f2911154e2feb8975621b78e379f12812 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrik=20Sj=C3=B6lin?=
 <44264371+ulrik-s@users.noreply.github.com>
Date: Sun, 29 Jun 2025 16:38:55 +0200
Subject: [PATCH 11/20] tests: fix broken chain in bblob fsck test

---
 t/t1030-bblob.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/t/t1030-bblob.sh b/t/t1030-bblob.sh
index 62827581307939..61285040b01430 100755
--- a/t/t1030-bblob.sh
+++ b/t/t1030-bblob.sh
@@ -39,7 +39,7 @@ test_expect_success 'server advertises bblob capability' '
 
 test_expect_success 'fsck verifies bblob objects' '
        git fsck --full-bblob-verify >out &&
-       ! grep "error" out || true
+       ! grep "error" out
 '
 
 test_done

From 07aabe9ef135f30833d0ab6669c93b3d09f558cf Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrik=20Sj=C3=B6lin?=
 <44264371+ulrik-s@users.noreply.github.com>
Date: Sun, 29 Jun 2025 21:13:01 +0200
Subject: [PATCH 12/20] tests: adjust capabilities expectations

---
 t/t5555-http-smart-common.sh |  1 +
 t/t5701-git-serve.sh         | 11 ++++++++---
 2 files changed, 9 insertions(+), 3 deletions(-)

diff --git a/t/t5555-http-smart-common.sh b/t/t5555-http-smart-common.sh
index e47ea1ad106048..fbea892d4872d9 100755
--- a/t/t5555-http-smart-common.sh
+++ b/t/t5555-http-smart-common.sh
@@ -130,6 +130,7 @@ test_expect_success 'git upload-pack --advertise-refs: v2' '
 	fetch=shallow wait-for-done
 	server-option
 	object-format=$(test_oid algo)
+	bblob
 	0000
 	EOF
 
diff --git a/t/t5701-git-serve.sh b/t/t5701-git-serve.sh
index d4c28bae39e2ad..969bc925497f99 100755
--- a/t/t5701-git-serve.sh
+++ b/t/t5701-git-serve.sh
@@ -35,7 +35,11 @@ test_expect_success 'setup to generate files with expected content' '
 '
 
 test_expect_success 'test capability advertisement' '
-	cat expect.base expect.trailer >expect &&
+       {
+               cat expect.base &&
+               echo bblob &&
+               cat expect.trailer
+       } >expect &&
 
 	if test_have_prereq WINDOWS
 	then
@@ -369,10 +373,11 @@ test_expect_success 'test capability advertisement with uploadpack.advertiseBund
 
 	cat >expect.extra <<-EOF &&
 	bundle-uri
+	bblob
 	EOF
 	cat expect.base \
-	    expect.extra \
-	    expect.trailer >expect &&
+            expect.extra \
+            expect.trailer >expect &&
 
 	if test_have_prereq WINDOWS
 	then

From d91c444583367e396e9727da72c17b6fe0b4ac77 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrik=20Sj=C3=B6lin?=
 <44264371+ulrik-s@users.noreply.github.com>
Date: Mon, 30 Jun 2025 05:08:40 +0200
Subject: [PATCH 13/20] tests: fix size comparison for bblob test

---
 t/t1030-bblob.sh | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/t/t1030-bblob.sh b/t/t1030-bblob.sh
index 61285040b01430..3c40626da40935 100755
--- a/t/t1030-bblob.sh
+++ b/t/t1030-bblob.sh
@@ -28,7 +28,7 @@ test_expect_success 'reading bblob yields original data' '
 '
 
 test_expect_success 'size helper matches original' '
-	test "$(git cat-file -s "$oid")" = "$(wc -c <bigfile)"
+       test $(git cat-file -s "$oid") -eq $(wc -c <bigfile)
 '
 
 test_expect_success 'server advertises bblob capability' '

From 08f8e88d904cda0274657b0ed59c80db4708a06f Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrik=20Sj=C3=B6lin?=
 <44264371+ulrik-s@users.noreply.github.com>
Date: Sat, 5 Jul 2025 07:28:28 +0300
Subject: [PATCH 14/20] tests: demonstrate fetch deduplication

---
 t/t5615-fetch-no-dup.sh | 26 ++++++++++++++++++++++++++
 1 file changed, 26 insertions(+)
 create mode 100755 t/t5615-fetch-no-dup.sh

diff --git a/t/t5615-fetch-no-dup.sh b/t/t5615-fetch-no-dup.sh
new file mode 100755
index 00000000000000..13d06b22356c13
--- /dev/null
+++ b/t/t5615-fetch-no-dup.sh
@@ -0,0 +1,26 @@
+#!/bin/sh
+
+test_description='fetch dedup'
+
+. ./test-lib.sh
+
+setup_repo() {
+    git init server &&
+    echo hello >server/file &&
+    (cd server && git add file && git commit -m initial)
+}
+
+setup_client() {
+    git clone server client
+}
+
+test_expect_success 'fetch from up-to-date repo is a no-op' '
+    setup_repo &&
+    setup_client &&
+    before=$(ls client/.git/objects/pack | wc -l) &&
+    git -C client fetch ../server >/dev/null 2>&1 &&
+    after=$(ls client/.git/objects/pack | wc -l) &&
+    test $before = $after
+'
+
+test_done

From 44b93971864efefbd527cec9e88e04231620ccfc Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrik=20Sj=C3=B6lin?=
 <44264371+ulrik-s@users.noreply.github.com>
Date: Sat, 5 Jul 2025 07:28:34 +0300
Subject: [PATCH 15/20] tests: extend fetch-no-dup coverage

---
 t/meson.build                                      | 1 +
 t/{t5615-fetch-no-dup.sh => t5622-fetch-no-dup.sh} | 0
 2 files changed, 1 insertion(+)
 rename t/{t5615-fetch-no-dup.sh => t5622-fetch-no-dup.sh} (100%)
 mode change 100755 => 100644

diff --git a/t/meson.build b/t/meson.build
index 78bd09419e2ea8..6914e85c5f70a9 100644
--- a/t/meson.build
+++ b/t/meson.build
@@ -726,6 +726,7 @@ integration_tests = [
   't5619-clone-local-ambiguous-transport.sh',
   't5620-backfill.sh',
   't5621-clone-revision.sh',
+  't5622-fetch-no-dup.sh',
   't5700-protocol-v1.sh',
   't5701-git-serve.sh',
   't5702-protocol-v2.sh',
diff --git a/t/t5615-fetch-no-dup.sh b/t/t5622-fetch-no-dup.sh
old mode 100755
new mode 100644
similarity index 100%
rename from t/t5615-fetch-no-dup.sh
rename to t/t5622-fetch-no-dup.sh

From f8794661b40ca1818b1d2cf39595b93bd97d8a55 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrik=20Sj=C3=B6lin?=
 <44264371+ulrik-s@users.noreply.github.com>
Date: Sat, 5 Jul 2025 07:28:38 +0300
Subject: [PATCH 16/20] upload-pack: parse bblob capability

---
 upload-pack.c | 23 +++++++++++++++--------
 1 file changed, 15 insertions(+), 8 deletions(-)

diff --git a/upload-pack.c b/upload-pack.c
index 8cc459d5f5cee0..7be8771ada8d2b 100644
--- a/upload-pack.c
+++ b/upload-pack.c
@@ -119,6 +119,7 @@ struct upload_pack_data {
 	unsigned allow_sideband_all : 1;			/* v2 only */
 	unsigned seen_haves : 1;				/* v2 only */
 	unsigned allow_packfile_uris : 1;			/* v2 only */
+	unsigned allow_bblob : 1;				/* v2 only */
 	unsigned advertise_sid : 1;
 	unsigned sent_capabilities : 1;
 };
@@ -154,6 +155,7 @@ static void upload_pack_data_init(struct upload_pack_data *data)
 
 	data->keepalive = 5;
 	data->advertise_sid = 0;
+	data->allow_bblob = 0;
 }
 
 static void upload_pack_data_clear(struct upload_pack_data *data)
@@ -1682,14 +1684,19 @@ static void process_args(struct packet_reader *request,
 			continue;
 		}
 
-		if (data->allow_packfile_uris &&
-		    skip_prefix(arg, "packfile-uris ", &p)) {
-			if (data->uri_protocols.nr)
-				send_err_and_die(data,
-						 "multiple packfile-uris lines forbidden");
-			string_list_split(&data->uri_protocols, p, ',', -1);
-			continue;
-		}
+               if (data->allow_packfile_uris &&
+                   skip_prefix(arg, "packfile-uris ", &p)) {
+                       if (data->uri_protocols.nr)
+                               send_err_and_die(data,
+                                                "multiple packfile-uris lines forbidden");
+                       string_list_split(&data->uri_protocols, p, ',', -1);
+                       continue;
+               }
+
+               if (!strcmp(arg, "bblob")) {
+                       data->allow_bblob = 1;
+                       continue;
+               }
 
 		/* ignore unknown lines maybe? */
 		die("unexpected line: '%s'", arg);

From 5d8c1bd5a223947ebc891b31dae5d29edec35611 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrik=20Sj=C3=B6lin?=
 <44264371+ulrik-s@users.noreply.github.com>
Date: Sat, 5 Jul 2025 07:28:43 +0300
Subject: [PATCH 17/20] fsck: validate bblob size more strictly

---
 fsck.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/fsck.c b/fsck.c
index 96c9f2a78028c9..57260f1198e28d 100644
--- a/fsck.c
+++ b/fsck.c
@@ -1259,9 +1259,9 @@ static int fsck_bblob(const struct object_id *oid, const char *buf,
            size_t oidsz = the_repository->hash_algo->rawsz;
            int i;
 
-	   if (size != oidsz * BBLOB_FANOUT)
-	       ret |= report(options, oid, OBJ_BBLOB, FSCK_MSG_BAD_TYPE,
-			      "invalid bblob size");
+       if (size % oidsz || size / oidsz > BBLOB_FANOUT)
+               ret |= report(options, oid, OBJ_BBLOB, FSCK_MSG_BAD_TYPE,
+                              "invalid bblob size");
 
            for (i = 0; i < BBLOB_FANOUT && i * oidsz < size; i++) {
                struct object_id child;

From 2687fa1bc685ea38adbb3e031afb6044e81f7340 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrik=20Sj=C3=B6lin?=
 <44264371+ulrik-s@users.noreply.github.com>
Date: Sat, 5 Jul 2025 07:28:47 +0300
Subject: [PATCH 18/20] bblob: use constants for chunk boundary

---
 bblob.c | 3 ++-
 bblob.h | 3 +++
 2 files changed, 5 insertions(+), 1 deletion(-)

diff --git a/bblob.c b/bblob.c
index 012fc3a5142d55..b0de03f03004af 100644
--- a/bblob.c
+++ b/bblob.c
@@ -80,7 +80,8 @@ int write_bblob(struct repository *r, const void *buf, unsigned long len,
                        git_hash_final(out, &c);
                        bits = (out[r->hash_algo->rawsz - 2] << 8) |
                                out[r->hash_algo->rawsz - 1];
-		       if ((bits & 0x1fff) == 0) {
+                       /* boundary when the trailing checksum hits a match */
+                       if ((bits & BBLOB_BREAK_MASK) == 0) {
                                struct object_id ch;
                                if (write_object_file((const char *)buf + chunk_start,
                                                     i - chunk_start + 1,
diff --git a/bblob.h b/bblob.h
index 370f6461c62205..eda8a156b2cb94 100644
--- a/bblob.h
+++ b/bblob.h
@@ -9,6 +9,9 @@
 /* heuristic target chunk size when splitting large blobs */
 #define BBLOB_CHUNK_GOAL 4096
 
+/* rolling checksum bits that must match to trigger a split */
+#define BBLOB_BREAK_MASK 0x1fff
+
 struct bblob {
        struct object object;
        struct object_id oids[BBLOB_FANOUT];

From 149c1e0f3581ef168f5ee17b03f4d23ab45bef61 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrik=20Sj=C3=B6lin?=
 <44264371+ulrik-s@users.noreply.github.com>
Date: Sat, 5 Jul 2025 07:28:52 +0300
Subject: [PATCH 19/20] tests: add smart-transfer bblob skeleton

---
 t/meson.build                   |  1 +
 t/t5623-bblob-smart-transfer.sh | 90 +++++++++++++++++++++++++++++++++
 2 files changed, 91 insertions(+)
 create mode 100755 t/t5623-bblob-smart-transfer.sh

diff --git a/t/meson.build b/t/meson.build
index 6914e85c5f70a9..065e7686aece51 100644
--- a/t/meson.build
+++ b/t/meson.build
@@ -727,6 +727,7 @@ integration_tests = [
   't5620-backfill.sh',
   't5621-clone-revision.sh',
   't5622-fetch-no-dup.sh',
+  't5623-bblob-smart-transfer.sh',
   't5700-protocol-v1.sh',
   't5701-git-serve.sh',
   't5702-protocol-v2.sh',
diff --git a/t/t5623-bblob-smart-transfer.sh b/t/t5623-bblob-smart-transfer.sh
new file mode 100755
index 00000000000000..bcd45408c71629
--- /dev/null
+++ b/t/t5623-bblob-smart-transfer.sh
@@ -0,0 +1,90 @@
+#!/bin/sh
+
+test_description='bblob smart transfer'
+
+. ./test-lib.sh
+
+cat_bigfile() {
+    perl -e "print \"$1\" x 20000" >bigfile
+}
+
+setup_server() {
+    git init server &&
+    (
+        cd server &&
+        cat_bigfile a &&
+        oid1=$(git hash-object -t bblob -w bigfile) &&
+        git update-index --add --cacheinfo 100644 $oid1 big.bin &&
+        git commit -m initial
+    )
+}
+
+setup_client() {
+    git clone server client &&
+    git -C client rev-parse HEAD > /dev/null
+}
+
+test_expect_success 'initial clone transfers bblob object' '
+    setup_server &&
+    setup_client &&
+    git -C server cat-file -p HEAD:big.bin >expect &&
+    git -C client cat-file -p HEAD:big.bin >actual &&
+    test_cmp expect actual
+'
+
+test_expect_failure 'fetch reuses existing bblob data' '
+    (
+        cd server &&
+        echo note >note &&
+        git add note &&
+        git commit -m second
+    ) &&
+    before=$(ls client/.git/objects/pack | wc -l) &&
+    git -C client fetch ../server >/dev/null &&
+    after=$(ls client/.git/objects/pack | wc -l) &&
+    test $((after-before)) = 1 &&
+    pack=$(ls client/.git/objects/pack/pack-*.pack | sort | tail -n1) &&
+    size=$(wc -c <"$pack") &&
+    test $size -lt 5000
+'
+
+test_expect_failure 'fetch transfers new bblob chunks only once' '
+    (
+        cd server &&
+        cat_bigfile b &&
+        oid2=$(git hash-object -t bblob -w bigfile) &&
+        git update-index --add --cacheinfo 100644 $oid2 big.bin &&
+        git commit -m third
+    ) &&
+    before=$(ls client/.git/objects/pack | wc -l) &&
+    git -C client fetch ../server >/dev/null &&
+    after=$(ls client/.git/objects/pack | wc -l) &&
+    test $((after-before)) = 1 &&
+    git -C server cat-file -p HEAD:big.bin >expect &&
+    git -C client cat-file -p FETCH_HEAD:big.bin >actual &&
+    test_cmp expect actual
+'
+
+test_expect_failure 'redundant fetch sends no additional pack' '
+    before=$(ls client/.git/objects/pack | wc -l) &&
+    git -C client fetch ../server >/dev/null &&
+    after=$(ls client/.git/objects/pack | wc -l) &&
+    test $before = $after
+'
+
+test_expect_failure 'reusing existing bblob avoids retransmission' '
+    (
+        cd server &&
+        git reset --hard HEAD^ &&
+        echo more >>note && git add note && git commit -m fourth
+    ) &&
+    before=$(ls client/.git/objects/pack | wc -l) &&
+    git -C client fetch ../server >/dev/null &&
+    after=$(ls client/.git/objects/pack | wc -l) &&
+    test $((after-before)) = 1 &&
+    pack=$(ls client/.git/objects/pack/pack-*.pack | sort | tail -n1) &&
+    size=$(wc -c <"$pack") &&
+    test $size -lt 5000
+'
+
+test_done

From bbf234b4aae970dbbf77131fc685be70597c432b Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Ulrik=20Sj=C3=B6lin?=
 <44264371+ulrik-s@users.noreply.github.com>
Date: Sat, 5 Jul 2025 07:28:56 +0300
Subject: [PATCH 20/20] auto-convert large blobs into bblobs

---
 object-file.c | 30 +++++++++++++++++++++++++++---
 1 file changed, 27 insertions(+), 3 deletions(-)

diff --git a/object-file.c b/object-file.c
index 880421977885f9..5cae7161df86c2 100644
--- a/object-file.c
+++ b/object-file.c
@@ -30,6 +30,10 @@
 #include "setup.h"
 #include "streaming.h"
 
+/* automatically convert large blobs to bblobs when writing */
+static int auto_bblob = 1;
+static int writing_bblob;
+
 /* The maximum size for an object header. */
 #define MAX_HEADER_LEN 32
 
@@ -1059,9 +1063,29 @@ int write_object_file_flags(const void *buf, unsigned long len,
 	struct repository *repo = the_repository;
 	const struct git_hash_algo *algo = repo->hash_algo;
 	const struct git_hash_algo *compat = repo->compat_hash_algo;
-	struct object_id compat_oid;
-	char hdr[MAX_HEADER_LEN];
-	int hdrlen = sizeof(hdr);
+       struct object_id compat_oid;
+       char hdr[MAX_HEADER_LEN];
+       int hdrlen = sizeof(hdr);
+
+       if (type == OBJ_BLOB && auto_bblob && !writing_bblob &&
+           len > BBLOB_CHUNK_GOAL) {
+               struct object_id bb;
+               writing_bblob = 1;
+               if (write_bblob(repo, buf, len, &bb)) {
+                       writing_bblob = 0;
+                       return -1;
+               }
+               writing_bblob = 0;
+               oidcpy(oid, &bb);
+               if (compat) {
+                       if (compat_oid_in)
+                               oidcpy(&compat_oid, compat_oid_in);
+                       else
+                               hash_object_file(compat, buf, len, OBJ_BLOB, &compat_oid);
+                       return repo_add_loose_object_map(repo, oid, &compat_oid);
+               }
+               return 0;
+       }
 
        /* Generate compat_oid */
        if (compat) {