This is a re-roll of [1] sent back in March. It removes the PPC_SHA1 implementation, which as noted in in 1/2 is likely completely unused at this point, if anyone's using it we have other better tested (and just as fast) SHA-1 implementations. I then included this change in a larger series sent in late April[2], which stalled for lack of feedback from the OSX crowd[3]. But this part should be uncontroversial, and is an obvious win in terms of the diffstat, and in reducing our complexity surface in this area. Changes since v2: * Much rewritten commit message (see range-diff) * Rebased on upstream changes * More PPC_SHA1 removal of comments/docs that I missed the first time around. * The s/C_OBJS/OBJECTS/ Makefile change is now split into a new 2/2. * Added brian's Acked-by, per [4] (not explicitly an "Acked-by", but I thought adding amounted to a fair paraphrasing of the feedback there) 1. https://lore.kernel.org/git/patch-v2-1.1-e77fd23a824-20220321T170412Z-avarab@xxxxxxxxx/ 2. https://lore.kernel.org/git/cover-0.5-00000000000-20220422T094624Z-avarab@xxxxxxxxx/ 3. https://lore.kernel.org/git/xmqqv8v17xrl.fsf@gitster.g/ 4. https://lore.kernel.org/git/Yjjr9fkybVmB53M7@xxxxxxxxxxxxxxxxxxxxxxxxx/ Ævar Arnfjörð Bjarmason (2): Makefile + hash.h: remove PPC_SHA1 implementation Makefile: use $(OBJECTS) instead of $(C_OBJ) INSTALL | 3 +- Makefile | 22 ++--- block-sha1/sha1.c | 4 - configure.ac | 3 - hash.h | 6 +- ppc/sha1.c | 72 --------------- ppc/sha1.h | 25 ------ ppc/sha1ppc.S | 224 ---------------------------------------------- 8 files changed, 9 insertions(+), 350 deletions(-) delete mode 100644 ppc/sha1.c delete mode 100644 ppc/sha1.h delete mode 100644 ppc/sha1ppc.S Range-diff against v2: 1: e77fd23a824 ! 1: 87a204b8937 ppc: remove custom SHA-1 implementation @@ Metadata Author: Ævar Arnfjörð Bjarmason <avarab@xxxxxxxxx> ## Commit message ## - ppc: remove custom SHA-1 implementation + Makefile + hash.h: remove PPC_SHA1 implementation Remove the PPC_SHA1 implementation added in a6ef3518f9a ([PATCH] PPC assembly implementation of SHA1, 2005-04-22). When this was added Apple consumer hardware used the PPC architecture, and the implementation was intended to improve SHA-1 speed there. - Since it was added we've moved to DC_SHA1 by default, and anyone - wanting hard-rolled non-DC SHA-1 implementation can use OpenSSL's via - the OPENSSL_SHA1 knob. + Since it was added we've moved to using sha1collisiondetection by + default, and anyone wanting hard-rolled non-DC SHA-1 implementation + can use OpenSSL's via the OPENSSL_SHA1 knob. - I'm unsure if this was ever supposed to work on 64-bit PPC. It clearly - originally targeted 32 bit PPC, but there's some mailing list - references to this being tried on G5 (PPC 970). I can't get it to do - anything but segfault on the BE POWER8 machine in the GCC compile - farm. Anyone caring about speed on PPC these days is likely to be - using IBM's POWER, not PPC 970. + The PPC_SHA1 originally originally targeted 32 bit PPC, and later the + 64 bit PPC 970 (a.k.a. Apple PowerPC G5). See 926172c5e48 (block-sha1: + improve code on large-register-set machines, 2009-08-10) for a + reference about the performance on G5 (a comment in block-sha1/sha1.c + being removed here). - There have been proposals to entirely remove non-DC_SHA1 - implementations from the tree[1]. I think per [2] that would be a bit + I can't get it to do anything but segfault on both the BE and LE POWER + machines in the GCC compile farm[1]. Anyone who's concerned about + performance on PPC these days is likely to be using the IBM POWER + processors. + + There have been proposals to entirely remove non-sha1collisiondetection + implementations from the tree[2]. I think per [3] that would be a bit overzealous. I.e. there are various set-ups git's speed is going to be more important than the relatively implausible SHA-1 collision attack, or where such attacks are entirely mitigated by other means (e.g. by incoming objects being checked with DC_SHA1). - The main reason for doing so at this point is to simplify follow-up - Makefile change. Since PPC_SHA1 included the only in-tree *.S assembly - file we needed to keep around special support for building objects - from it. By getting rid of it we know we'll always build *.o from *.c - files, which makes the build process simpler. + But that really doesn't apply to PPC_SHA1 in particular, which seems + to have outlived its usefulness. + + As this gets rid of the only in-tree *.S assembly file we can remove + the small bits of logic from the Makefile needed to build objects + from *.S (as opposed to *.c) - As an aside the code being removed here was also throwing warnings - with the "-pedantic" flag, but let's remove it instead of fixing it, - as 544d93bc3b4 (block-sha1: remove use of obsolete x86 assembly, - 2022-03-10) did for block-sha1/*. + The code being removed here was also throwing warnings with the + "-pedantic" flag, it could have been fixed as 544d93bc3b4 (block-sha1: + remove use of obsolete x86 assembly, 2022-03-10) did for block-sha1/*, + but as noted above let's remove it instead. - 1. https://lore.kernel.org/git/20200223223758.120941-1-mh@xxxxxxxxxxxx/ - 2. https://lore.kernel.org/git/20200224044732.GK1018190@xxxxxxxxxxxxxxxxxxxxxxx/ + 1. https://cfarm.tetaneutral.net/machines/list/ + Tested on gcc{110,112,135,203}, a mixture of POWER [789] ppc64 and + ppc64le. All segfault in anything needing object + hashing (e.g. t/t1007-hash-object.sh) when compiled with + PPC_SHA1=Y. + 2. https://lore.kernel.org/git/20200223223758.120941-1-mh@xxxxxxxxxxxx/ + 3. https://lore.kernel.org/git/20200224044732.GK1018190@xxxxxxxxxxxxxxxxxxxxxxx/ + Acked-by: brian m. carlson" <sandals@xxxxxxxxxxxxxxxxxxxx> Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@xxxxxxxxx> ## INSTALL ## @@ Makefile: include shared.mak # Define DC_SHA1 to unconditionally enable the collision-detecting sha1 # algorithm. This is slower, but may detect attempted collision attacks. # Takes priority over other *_SHA1 knobs. -@@ Makefile: ifdef OPENSSL_SHA1 - EXTLIBS += $(LIB_4_CRYPTO) - BASIC_CFLAGS += -DSHA1_OPENSSL - else +@@ Makefile: ifdef APPLE_COMMON_CRYPTO + SHA1_MAX_BLOCK_SIZE = 1024L*1024L*1024L + endif + +ifdef PPC_SHA1 -+$(error PPC_SHA1 has been removed! Use DC_SHA1 instead, which is the default) ++$(error the PPC_SHA1 flag has been removed along with the PowerPC-specific SHA-1 implementation.) +endif - ifdef BLK_SHA1 ++ + ifdef OPENSSL_SHA1 + EXTLIBS += $(LIB_4_CRYPTO) + BASIC_CFLAGS += -DSHA1_OPENSSL +@@ Makefile: ifdef BLK_SHA1 LIB_OBJS += block-sha1/sha1.o BASIC_CFLAGS += -DSHA1_BLK else @@ Makefile: missing_compdb_dir = -ASM_SRC := $(wildcard $(OBJECTS:o=S)) -ASM_OBJ := $(ASM_SRC:S=o) -C_OBJ := $(filter-out $(ASM_OBJ),$(OBJECTS)) -+C_OBJ = $(OBJECTS) ++C_OBJ := $(OBJECTS) $(C_OBJ): %.o: %.c GIT-CFLAGS $(missing_dep_dirs) $(missing_compdb_dir) $(QUIET_CC)$(CC) -o $*.o -c $(dep_args) $(compdb_args) $(ALL_CFLAGS) $(EXTRA_CPPFLAGS) $< @@ Makefile: missing_compdb_dir = %.s: %.c GIT-CFLAGS FORCE $(QUIET_CC)$(CC) -o $@ -S $(ALL_CFLAGS) $(EXTRA_CPPFLAGS) $< + ## block-sha1/sha1.c ## +@@ + * try to do the silly "optimize away loads" part because it won't + * see what the value will be). + * +- * Ben Herrenschmidt reports that on PPC, the C version comes close +- * to the optimized asm with this (ie on PPC you don't want that +- * 'volatile', since there are lots of registers). +- * + * On ARM we get the best code generation by forcing a full memory barrier + * between each SHA_ROUND, otherwise gcc happily get wild with spilling and + * the stack frame size simply explode and performance goes down the drain. + ## configure.ac ## @@ configure.ac: AC_MSG_NOTICE([CHECKS for site configuration]) # tests. These tests take up a significant amount of the total test time -: ----------- > 2: cb3bc8b5029 Makefile: use $(OBJECTS) instead of $(C_OBJ) -- 2.37.3.1406.g184357183a6
