Re: PR/51569 CVS commit: src/crypto/external/bsd/openssl/dist/crypto/bn/asm

[Andreas Gustafsson <gson%gson.org@localhost>]

The following reply was made to PR lib/51569; it has been noted by GNATS.

From: Andreas Gustafsson <gson%gson.org@localhost>
To: christos%NetBSD.org@localhost
Cc: gnats-bugs%netbsd.org@localhost, coypu%SDF.ORG@localhost, spz%NetBSD.org@localhost
Subject: Re: PR/51569 CVS commit: src/crypto/external/bsd/openssl/dist/crypto/bn/asm
Date: Mon, 9 Jan 2017 17:46:27 +0200

 Christos,
 
 You wrote:
 >  Log Message:
 >  PR/51569: Andreas Gustafsson: Check the right bit for pclmulqdq:
 >  Perform a Carry-Less Multiplication of Quadword instruction
 >  (accelerator for GCM)
 >  
 >  To generate a diff of this commit:
 >  cvs rdiff -u -r1.4 -r1.5 \
 >      src/crypto/external/bsd/openssl/dist/crypto/bn/asm/x86_64-gf2m.pl
 
 I think the actual root cause of this bug was the line
 
   -e 's/\(OPENSSL[A-Za-z0-9_]*\)(%rip)/\1@GOTPCREL(%rip)/' \
 
 in src/crypto/external/bsd/openssl/lib/libcrypto/arch/x86_64/Makefile,
 which you later removed in revision 1.11.
 
 I think the change to x86_64-gf2m.pl worked simply because the added
 "+4" caused the regexp not to match, so that the @GOTPCREL was not
 added.  Other than that, it seems to me the old and new code should
 give the same result (i.e., test the same bit).  Now that the
 @GOTPCREL hack is gone from the Makefile, there should no longer be
 any need for the change to x86_64-gf2m.pl, so I think it should be
 reverted in the interest of keeping the differences wrt upstream
 minimal.
 
 I have a local change to revert the x86_64-gf2m.pl change - OK to
 commit that and the corresponding regenerated .S file?
 -- 
 Andreas Gustafsson, gson%gson.org@localhost