Need CAN-2004-0079 & CAN-2004-0112 patch for testing

[openssl.org #836] SSL_CTX_set_cipher_list() in OpenSSL 0.9.7c under Linux

        I am working with OpenSSL 0.9.7c under Linux. I have come across
the following when using SSL_CTX_set_cipher_list() in this platform.

        According to man page, SSL_CTX_set_cipher_list() has the
following prototype:

        int SSL_CTX_set_cipher_list(SSL_CTX *ctx, const char *str)

The str argument is a string that identifies the cipher suite to be
selected. Again according to the man page, quote, failure is only
flagged if no ciphers could be collected at all, unquote. On this basis
I would have thought that using some random string for str should
cause SSL_CTX_set_cipher_list() to return a failure. Apparently, it
doesn't.

        Looking into the code, SSL_CTX_set_cipher_list() depends on 
ssl_create_cipher_list() for all the actual work. If
ssl_create_cipher_list() returns a pointer to a valid
STACK_OF(SSL_CIPHER) then SSL_CTX_set_cipher_list() will return 1
(success) otherwise it will return 0 (failure).

        For concreteness let's choose a specific random string:
"abcdef". When SSL_CTX_set_cipher_list() is invoked on this string (I'll
assume henceforth that the ctx argument is fine),
ssl_create_cipher_list() returns a non-NULL pointer:

(gdb) p *sk

$2 = {
  num = 0, 
  data = 0x6000000000032270,
  sorted = 0, 
  num_alloc = 4, 
  comp = 0
}

        At 0x6000000000032270 we only find strings of zeros - i.e. no
cipher suite-related data. 

	If we use the "DES-CBC-SHA" string instead as the value of str, then
ssl_create_cipher_list() returns a non-NULL pointer again, but this time
with useful data in the address at the data field. In addition, the num
field is 1, rather than 0, as above.

        It is clear that for "abcdef" ssl_create_cipher_list(), as
expected, has failed to recognize a valid cipher suite, and therefore
the value returned by SSL_CTX_set_cipher_list() ought to be 0, not 1.
However, since SSL_CTX_set_cipher_list() relies on the NULL or non-NULL
character of the pointer returned by ssl_create_cipher_list(), it
returns 1 for this nonsensical "abcdef" string. 

        While this is not a big technical deal (the goodness or not of
the STACK_OF(SSL_CIPHER) pointer can be determined by looking at the num
field, I guess) I believe that this behavior is not in agreement with
the description in the manual page for SSL_CTX_set_cipher_list(). In my
view, this function should be changed to read as follows:

/** specify the ciphers to be used by default by the SSL_CTX */
int SSL_CTX_set_cipher_list(SSL_CTX *ctx, const char *str)
        {
        STACK_OF(SSL_CIPHER) *sk;

        sk=ssl_create_cipher_list(ctx->method,&ctx->cipher_list,
                &ctx->cipher_list_by_id,str);

/* XXXX */
        return(((sk == NULL) || (sk->num == 0)) ? 0 : 1);
        }

	Cheers, 

			JCA

______________________________________________________________________
OpenSSL Project                                 http://www.openssl.org
Development Mailing List                       openssl-dev <at> openssl.org
Automated List Manager                           majordomo <at> openssl.org

[openssl.org #837] Bugs in ocsp_vfy.c

Hello -

We're coding against OpenSSL 0.9.7c and discovered the following bugs. 
Both bugs occur in the the file crypto/ocsp/ocsp_vfy.c, and are similar 
in nature.  They are found in loops found in the methods ocsp_check_ids 
and ocsp_match_issuerid, and are as follows:

ocsp_check_ids:

for (i = 1; i < idcount; i++) {
   tmpid = sk_OCSP_SINGLERESP_value(sresp, 0)->certId;
   ... code removed ...
}

ocsp_match_issuerid:

for (i = 0; i < sk_OCSP_SINGLERESP_num(sresp); i++) {
   tmpid = sk_OCSP_SINGLERESP_value(sresp, 0)->certId;
   ... code removed ...
}

You can see the first line in each of these loops accesses the ocsp 
single response object not at index i, but at index 0 (making the loop 
redundant).  We believe that line should be changed to appear as follows:

tmpid = sk_OCSP_SINGLERESP_value(sresp, i)->certId;

Please let me know if there is a more appropriate means of reporting 
this bug, or if any further clarification is needed.  Thanks.

Paul Siegel
______________________________________________________________________
OpenSSL Project                                 http://www.openssl.org
Development Mailing List                       openssl-dev <at> openssl.org
Automated List Manager                           majordomo <at> openssl.org

[openssl.org #837] Bugs in ocsp_vfy.c

[psiegel <at> corestreet.com - Wed Mar  3 18:38:37 2004]:

> Hello -
> 
> We're coding against OpenSSL 0.9.7c and discovered the following bugs.
> Both bugs occur in the the file crypto/ocsp/ocsp_vfy.c, and are
> similar
> in nature.  They are found in loops found in the methods
> ocsp_check_ids
> and ocsp_match_issuerid, and are as follows:
> 
> ocsp_check_ids:
> 
> for (i = 1; i < idcount; i++) {
>    tmpid = sk_OCSP_SINGLERESP_value(sresp, 0)->certId;
>    ... code removed ...
> }
> 
> ocsp_match_issuerid:
> 
> for (i = 0; i < sk_OCSP_SINGLERESP_num(sresp); i++) {
>    tmpid = sk_OCSP_SINGLERESP_value(sresp, 0)->certId;
>    ... code removed ...
> }
> 
> You can see the first line in each of these loops accesses the ocsp
> single response object not at index i, but at index 0 (making the loop
> redundant).  We believe that line should be changed to appear as
> follows:
> 
> tmpid = sk_OCSP_SINGLERESP_value(sresp, i)->certId;
> 
> Please let me know if there is a more appropriate means of reporting
> this bug, or if any further clarification is needed.  Thanks.
> 

Yes you are correct. Thanks for the report.

Its probably because few OCSP reponses contain more than one certificate
details and you'd only notice during an error condition that this wasn't
spotted before.

I'll check in a fix.

Steve.

______________________________________________________________________
OpenSSL Project                                 http://www.openssl.org
Development Mailing List                       openssl-dev <at> openssl.org
Automated List Manager                           majordomo <at> openssl.org

Re: [openssl.org #832] ocsp and dsa key+socket option SO_REUSEADDR for responder

evp_md = EVP_get_digestbynid(rkey->type) return:
- DSA key: same as EVP_dss1()
- RSA key: NULL - OBJ_nid2sn(rkey->type) return "rsaEncryption" 

Richard Levitte - VMS Whacker wrote:

>In message <rt-832-5266.0.7645538094242 <at> openssl.org> on Fri, 27 Feb 2004 14:00:24 +0100 (MET),
"Stephen Henson via RT" <rt <at> openssl.org> said:
>
>rt> 
>rt> [levitte - Fri Feb 27 00:04:45 2004]:
>rt> 
>rt> > I noticed the EVP_dss1() hack, and am pondering about something a bit 
>rt> > mkore generic.  Could you test if the following line works?
>rt> > 
>rt> >   evp_md = EVP_get_digestbynid(EVP_MD_nid(rkey));
>rt> > 
>rt> > [guest - Mon Feb 23 18:13:00 2004]:
>rt> > 
>rt> > > Attached file extend OCSP responder with:
>rt> > > - add support for dsa keys;
>rt> > > - set SO_REUSEADDR when system support it.
>rt> > > 
>rt> > > What about attached patch ?
>rt> > > It is against 0.9.7b, but same fixes can be applied for 0.9.7c.
>rt> > 
>rt> 
>rt> Its only because both EVP_PKEY and EVP_MD structures both have a field
>rt> called "type" and that EVP_MD_nid() is a macro that it works.
>rt> 
>rt> If ever EVP_MD_nid() was changed to a function bad things could happen.
>[...]
>rt> The applications currently access pkey->type directly for this kind of
>rt> thing.
>
>So basically, you're saying that the following would be the right
>thing to do:
>
>    evp_md = EVP_get_digestbynid(rkey->type);
>
>-----
>Please consider sponsoring my work on free software.
>See http://www.free.lp.se/sponsoring.html for details.
>
>  
>

______________________________________________________________________
OpenSSL Project                                 http://www.openssl.org
Development Mailing List                       openssl-dev <at> openssl.org
Automated List Manager                           majordomo <at> openssl.org

SMCard

Dear Group
   I buy a SmartCard. I want to load a private key to
it.I get a module from seller that the related
function to load the private key need some
parameters.I don't know what's this parameters and how
I must set these parameters?
The following are declaration of that function:

WORD SetPrivateKey( BYTE i_byKeyNo , BYTE *i_pbyDPDQ ,
BYTE *i_pbyPQU );
    20    // i_byKeyNo: Key number: 0..4
    21    // i_pbyDPDQ: A pointer to DP and DQ part of
private key (Chinese Remainder ???
    22    // i_pbyPQU:  A pointer to P,Q, and U
???part of private key (Chinese Remainder ???
    23    // Output:    0x9000 -> Successful , Other
-> Unsuccessful
I review the RSA structure and found something for
example DP is equivalent to dmp1 and DQ is equivalent
to dmq1 and P is equivalent to p and Q is equivalent
to q. But I''m confused that how we can have a pointer
to point to two pointer in the i_pbyDPDQ.

typedef struct rsa_st
    86    {
    87    /* The first parameter is used to pickup
errors where
    88     * this is passed instead of aEVP_PKEY, it
is set to 0 */
    89    int pad;
    90    int version;
    91    RSA_METHOD *meth;
    92    BIGNUM *n;
    93    BIGNUM *e;
    94    BIGNUM *d;
    95    BIGNUM *p;
    96    BIGNUM *q;
    97    BIGNUM *dmp1;
    98    BIGNUM *dmq1;
    99    BIGNUM *iqmp;
   100    /* be carefull using this if the RSA
structure is shared */
   101    CRYPTO_EX_DATA ex_data;
   102    int references;
   103    int flags;

Please help me.
Bye.

__________________________________
Do you Yahoo!?
Yahoo! Search - Find what you’re looking for faster
http://search.yahoo.com
______________________________________________________________________
OpenSSL Project                                 http://www.openssl.org
Development Mailing List                       openssl-dev <at> openssl.org
Automated List Manager                           majordomo <at> openssl.org

Re: [openssl.org #788] sse2 bn_mul_add_words: tsu notification - encryption

this is the long overdue TSU notification for my patch to enhance
bn_mul_add_words with SSE2 support.

unfortunately i never got the time to make the various improvements
suggested on the mailing list, but i'm hoping that by clearing up the
export requirements on this patch someone else will be able to clean
things up for final inclusion.

thanks
-dean

SUBMISSION TYPE: TSU
SUBMITTED BY: dean gaudet
SUBMITTED FOR: dean gaudet
POINT OF CONTACT: dean-openssl <at> arctic.org
PHONE and/or FAX: 408 919 3086
MANUFACTURER: n/a
PRODUCT NAME/MODEL #: openssl
ECCN: 5D002

this patch introduces dynamic selection of the bn_mul_add_words inner
loop according to whether the processor supports SSE2 or not.

	./config CPUID

to enable this support.  (this is all that's required to get the SSE2
support.)

the cpuid.c file definitely needs porting to windoze (due to the need
to trap SIGILL during the test).  i've tested it on recent linux,
freebsd and netbsd; and on p4, xeon, athlon, opteron, pentium-m, p3,
p3m, and efficeon.

i'm not really sure where to cause my CRYPTO_cpuid_init() routine
to be called.  i've just tacked in a couple calls here in BN_new()
and BN_init() ... but it needs some place more central.

see http://arctic.org/~dean/crypto/rsa.html for the latest info.

-dean

===================================================================
RCS file: /home/dean/openssl/Repository/openssl/Configure,v
retrieving revision 1.419
diff -u -r1.419 Configure
--- Configure.orig	30 Nov 2003 23:29:26 -0000	1.419
+++ Configure	8 Dec 2003 03:24:48 -0000
 <at>  <at>  -90,6 +90,7  <at>  <at> 
 # MD5_ASM	use some extra md5 assember,
 # SHA1_ASM	use some extra sha1 assember, must define L_ENDIAN for x86
 # RMD160_ASM	use some extra ripemd160 assember,
+# CPUID		use x86 cpuid to select MMX/SSE2 code at run-time

 my $x86_gcc_des="DES_PTR DES_RISC1 DES_UNROLL";

 <at>  <at>  -636,6 +637,7  <at>  <at> 
 my $sha1_obj="";
 my $rmd160_obj="";
 my $processor="";
+my $cpuid=0;
 my $default_ranlib;
 my $perl;

 <at>  <at>  -832,6 +834,8  <at>  <at> 
 			}
 		elsif (/^386$/)
 			{ $processor=386; }
+		elsif (/^CPUID$/)
+			{ $cpuid=1; }
 		elsif (/^rsaref$/)
 			{
 			# No RSAref support any more since it's not needed.
 <at>  <at>  -1200,6 +1204,7  <at>  <at> 
 #	$rmd160_obj=$rmd160_enc;
 	$cflags.=" -DRMD160_ASM";
 	}
+$cflags.=" -DCPUID" if ($cpuid);

 # "Stringify" the C flags string.  This permits it to be made part of a string
 # and works as well on command lines.
===================================================================
RCS file: /home/dean/openssl/Repository/openssl/crypto/Makefile.ssl,v
retrieving revision 1.101
diff -u -r1.101 Makefile.ssl
--- crypto/Makefile.ssl.orig	3 Dec 2003 16:29:41 -0000	1.101
+++ crypto/Makefile.ssl	8 Dec 2003 04:01:37 -0000
 <at>  <at>  -38,8 +38,8  <at>  <at> 

 LIB= $(TOP)/libcrypto.a
 SHARED_LIB= libcrypto$(SHLIB_EXT)
-LIBSRC=	cryptlib.c mem.c mem_clr.c mem_dbg.c cversion.c ex_data.c tmdiff.c cpt_err.c ebcdic.c
uid.c o_time.c o_str.c
-LIBOBJ= cryptlib.o mem.o mem_clr.o mem_dbg.o cversion.o ex_data.o tmdiff.o cpt_err.o ebcdic.o uid.o
o_time.o o_str.o
+LIBSRC=	cryptlib.c mem.c mem_clr.c mem_dbg.c cversion.c ex_data.c tmdiff.c cpt_err.c ebcdic.c
uid.c o_time.c o_str.c cpuid.c
+LIBOBJ= cryptlib.o mem.o mem_clr.o mem_dbg.o cversion.o ex_data.o tmdiff.o cpt_err.o ebcdic.o uid.o
o_time.o o_str.o cpuid.o

 SRC= $(LIBSRC)

===================================================================
RCS file: crypto/cpuid.c
diff -N crypto/cpuid.c
--- crypto/cpuid.c.orig	1 Jan 1970 00:00:00 -0000
+++ crypto/cpuid.c	8 Dec 2003 05:56:16 -0000
 <at>  <at>  -0,0 +1,158  <at>  <at> 
+/*
+ * This x86 CPUID code is derived from intel appnote 241618.
+ *
+ * This code is public domain.
+ */
+
+/*
+ * Unfortunately, this code isn't portable.  Naively it would
+ * seem like this stuff belongs in one of the perlasm files,
+ * however there's a nasty SSE detection problem:  you
+ * need to execute a possibly faulting instruction in order
+ * to distinguish the setting of cr4.osfxsr on CPUs which
+ * claim to support SSE.  If the OS kernel has not set
+ * cr4.osfxsr=1 (to indicate that it will save/restore the
+ * extra registers via FXSAVE/FXRESTOR on context switches)
+ * then SSE/SSE2 instructions will fault, even though cpuid
+ * indicates their availability.
+ *
+ * Intercepting the fault requires OS interfacing... on unix
+ * it requires trapping SIGILL.  On Win32, I'm not sure what
+ * to do.  At any rate, it would be a royal pain to do all
+ * the signal (and threadsafe) foo from assembly, so this
+ * code is in C with embedded asm.
+ *
+ * - dean <at> arctic.org
+ */
+
+#ifdef CPUID
+#include <stdlib.h>
+#include <signal.h>
+#include <setjmp.h>
+#ifdef _REENTRANT
+#include <pthread.h>
+#endif
+
+#include "cpuid.h"
+
+unsigned CRYPTO_cpuid_value;
+
+
+static sigjmp_buf CRYPTO_cpuid_jmp;
+
+static void CRYPTO_cpuid_sigill(int signum)
+{
+	siglongjmp(CRYPTO_cpuid_jmp, 1);
+}
+
+static unsigned CRYPTO_cpuid_init_internal(void)
+{
+	char *e;
+	unsigned eax, ebx, ecx, edx;
+
+	CRYPTO_cpuid_value = CRYPTO_CPUID_ONCE;
+
+	e = getenv("CRYPTO_CPUID");
+	if (e) {
+		CRYPTO_cpuid_value = strtoul(e, 0, 0) | CRYPTO_CPUID_ONCE;
+		return CRYPTO_cpuid_value;
+	}
+
+	/* test for the existance of the CPUID instruction
+	 * by toggling the ID bit in eflags register.
+	 * see intel application note 241618
+	 */
+	__asm volatile(
+		"\n	pushf"
+		"\n	pop %0"
+		"\n	mov %0,%1"
+		"\n	xor $0x200000,%0"
+		"\n	push %0"
+		"\n	popf"
+		"\n	pushf"
+		"\n	pop %0"
+		"\n	xor %0,%1"
+		"\n	mov $0,%0"
+		"\n	jz 1f"
+		"\n	mov $1,%0"
+		"\n1:"
+		: "=&r" (eax), "=&r" (ecx) :: "memory");
+	if (!eax) {
+		return CRYPTO_cpuid_value;
+	}
+
+	/* get results of cpuid level 1 */
+	__asm volatile(
+		"cpuid"
+		: "=a" (eax), "=b" (ebx), "=c" (ecx), "=d" (edx)
+		: "0" (1));
+
+	if (edx & 0x800000) {
+		CRYPTO_cpuid_value |= CRYPTO_CPUID_MMX;
+	}
+
+	/* here is the extended test for SSE which checks if an
+	 * SSE instruction traps.
+	 */
+	if (edx & 0x2000000) {
+		struct sigaction sa, sa_save;
+
+		if (sigsetjmp(CRYPTO_cpuid_jmp, 1) == 0) {
+			sa.sa_handler = CRYPTO_cpuid_sigill;
+			sa.sa_flags = 0;
+			sigemptyset(&sa.sa_mask);
+			if (sigaction(SIGILL, &sa, &sa_save)) {
+				return CRYPTO_cpuid_value;
+			}
+
+			/* this is "xorps %xmm0,%xmm0" ... in a manner
+			 * which all as(1) should accept.
+			 */
+			__asm volatile(".byte 0x0f,0x57,0xc0");
+		}
+		else {
+			sigaction(SIGILL, &sa_save, NULL);
+			return CRYPTO_cpuid_value;
+		}
+		sigaction(SIGILL, &sa_save, NULL);
+
+		CRYPTO_cpuid_value |= CRYPTO_CPUID_SSE;
+	}
+
+	if (edx & 0x4000000) {
+		CRYPTO_cpuid_value |= CRYPTO_CPUID_SSE2;
+	}
+
+	if (ecx & 1) {
+		CRYPTO_cpuid_value |= CRYPTO_CPUID_SSE3;
+	}
+
+	return CRYPTO_cpuid_value;
+}
+
+
+#ifdef _REENTRANT
+static pthread_once_t CRYPTO_cpuid_once_control = PTHREAD_ONCE_INIT;
+
+unsigned CRYPTO_cpuid_init(void)
+{
+	if (CRYPTO_cpuid_value) {
+		return CRYPTO_cpuid_value;
+	}
+	pthread_once(&CRYPTO_cpuid_once_control,
+			(void (*)(void))CRYPTO_cpuid_init_internal);
+	return CRYPTO_cpuid_value;
+}
+#else
+unsigned CRYPTO_cpuid_init(void)
+{
+	if (CRYPTO_cpuid_value) {
+		return CRYPTO_cpuid_value;
+	}
+	return CRYPTO_cpuid_init_internal();
+}
+#endif
+
+#else
+int CPUID_dummy;
+#endif
===================================================================
RCS file: crypto/cpuid.h
diff -N crypto/cpuid.h
--- crypto/cpuid.h.orig	1 Jan 1970 00:00:00 -0000
+++ crypto/cpuid.h	8 Dec 2003 05:21:57 -0000
 <at>  <at>  -0,0 +1,28  <at>  <at> 
+#ifndef HEADER_CPUID_H
+#define HEADER_CPUID_H
+
+#ifdef  __cplusplus
+extern "C" {
+#endif
+
+#define CRYPTO_CPUID_ONCE		(0x01)
+#define CRYPTO_CPUID_MMX		(0x02)
+#define CRYPTO_CPUID_SSE		(0x04)
+#define CRYPTO_CPUID_SSE2		(0x08)
+#define CRYPTO_CPUID_SSE3		(0x10)
+
+#ifdef CPUID
+extern unsigned CRYPTO_cpuid_init(void);
+extern unsigned CRYPTO_cpuid_value;
+
+#define CRYPTO_cpuid()	(CRYPTO_cpuid_value ? CRYPTO_cpuid_value : CRYPTO_cpuid_init())
+
+#else
+#define CRYPTO_cpuid()	(0)
+#endif
+
+#ifdef  __cplusplus
+}
+#endif
+
+#endif
===================================================================
RCS file: /home/dean/openssl/Repository/openssl/crypto/bn/bn_lib.c,v
retrieving revision 1.55
diff -u -r1.55 bn_lib.c
--- crypto/bn/bn_lib.c.orig	2 Dec 2003 20:01:30 -0000	1.55
+++ crypto/bn/bn_lib.c	8 Dec 2003 04:25:01 -0000
 <at>  <at>  -66,6 +66,7  <at>  <at> 
 #include <stdio.h>
 #include "cryptlib.h"
 #include "bn_lcl.h"
+#include "cpuid.h"

 const char *BN_version="Big Number" OPENSSL_VERSION_PTEXT;

 <at>  <at>  -288,6 +289,7  <at>  <at> 

 void BN_init(BIGNUM *a)
 	{
+	(void)CRYPTO_cpuid();
 	memset(a,0,sizeof(BIGNUM));
 	bn_check_top(a);
 	}
 <at>  <at>  -296,6 +298,7  <at>  <at> 
 	{
 	BIGNUM *ret;

+	(void)CRYPTO_cpuid();
 	if ((ret=(BIGNUM *)OPENSSL_malloc(sizeof(BIGNUM))) == NULL)
 		{
 		BNerr(BN_F_BN_NEW,ERR_R_MALLOC_FAILURE);
===================================================================
RCS file: /home/dean/openssl/Repository/openssl/crypto/bn/asm/bn-586.pl,v
retrieving revision 1.5
diff -u -r1.5 bn-586.pl
--- crypto/bn/asm/bn-586.pl.orig	6 Dec 2000 16:30:23 -0000	1.5
+++ crypto/bn/asm/bn-586.pl	8 Dec 2003 04:45:43 -0000
 <at>  <at>  -42,6 +42,82  <at>  <at> 

 	&jz(&label("maw_finish"));

+	if ($cpuid) {
+		&test(&DWP("CRYPTO_cpuid_value","","",0),$cpuid_sse2);
+		&jz(&label("maw_loop"));
+
+		&movd("mm0",$w);		# mm0 = w
+		&pxor("mm1","mm1");		# mm1 = carry_in
+
+		&set_label("maw_sse2_loop",0);
+		&movd("mm3",&DWP(0,$r,"",0));	# mm3 = r[0]
+		&paddq("mm1","mm3");		# mm1 = carry_in + r[0]
+		&movd("mm2",&DWP(0,$a,"",0));	# mm2 = a[0]
+		&pmuludq("mm2","mm0");		# mm2 = w*a[0]
+		&movd("mm4",&DWP(4,$a,"",0));	# mm4 = a[1]
+		&pmuludq("mm4","mm0");		# mm4 = w*a[1]
+		&movd("mm6",&DWP(8,$a,"",0));	# mm6 = a[2]
+		&pmuludq("mm6","mm0");		# mm6 = w*a[2]
+		&movd("mm7",&DWP(12,$a,"",0));	# mm7 = a[3]
+		&pmuludq("mm7","mm0");		# mm7 = w*a[3]
+		&paddq("mm1","mm2");		# mm1 = carry_in + r[0] + w*a[0]
+		&movd("mm3",&DWP(4,$r,"",0));	# mm3 = r[1]
+		&paddq("mm3","mm4");		# mm3 = r[1] + w*a[1]
+		&movd("mm5",&DWP(8,$r,"",0));	# mm5 = r[2]
+		&paddq("mm5","mm6");		# mm5 = r[2] + w*a[2]
+		&movd("mm4",&DWP(12,$r,"",0));	# mm4 = r[3]
+		&paddq("mm7","mm4");		# mm7 = r[3] + w*a[3]
+		&movd(&DWP(0,$r,"",0),"mm1");
+		&movd("mm2",&DWP(16,$a,"",0));	# mm2 = a[4]
+		&pmuludq("mm2","mm0");		# mm2 = w*a[4]
+		&psrlq("mm1",32);		# mm1 = carry0
+		&movd("mm4",&DWP(20,$a,"",0));	# mm4 = a[5]
+		&pmuludq("mm4","mm0");		# mm4 = w*a[5]
+		&paddq("mm1","mm3");		# mm1 = carry0 + r[1] + w*a[1]
+		&movd("mm6",&DWP(24,$a,"",0));	# mm6 = a[6]
+		&pmuludq("mm6","mm0");		# mm6 = w*a[6]
+		&movd(&DWP(4,$r,"",0),"mm1");
+		&psrlq("mm1",32);		# mm1 = carry1
+		&movd("mm3",&DWP(28,$a,"",0));	# mm3 = a[7]
+		&add($a,32);
+		&pmuludq("mm3","mm0");		# mm3 = w*a[7]
+		&paddq("mm1","mm5");		# mm1 = carry1 + r[2] + w*a[2]
+		&movd("mm5",&DWP(16,$r,"",0));	# mm5 = r[4]
+		&paddq("mm2","mm5");		# mm2 = r[4] + w*a[4]
+		&movd(&DWP(8,$r,"",0),"mm1");
+		&psrlq("mm1",32);		# mm1 = carry2
+		&paddq("mm1","mm7");		# mm1 = carry2 + r[3] + w*a[3]
+		&movd("mm5",&DWP(20,$r,"",0));	# mm5 = r[5]
+		&paddq("mm4","mm5");		# mm4 = r[5] + w*a[5]
+		&movd(&DWP(12,$r,"",0),"mm1");
+		&psrlq("mm1",32);		# mm1 = carry3
+		&paddq("mm1","mm2");		# mm1 = carry3 + r[4] + w*a[4]
+		&movd("mm5",&DWP(24,$r,"",0));	# mm5 = r[6]
+		&paddq("mm6","mm5");		# mm6 = r[6] + w*a[6]
+		&movd(&DWP(16,$r,"",0),"mm1");
+		&psrlq("mm1",32);		# mm1 = carry4
+		&paddq("mm1","mm4");		# mm1 = carry4 + r[5] + w*a[5]
+		&movd("mm5",&DWP(28,$r,"",0));	# mm5 = r[7]
+		&paddq("mm3","mm5");		# mm3 = r[7] + w*a[7]
+		&movd(&DWP(20,$r,"",0),"mm1");
+		&psrlq("mm1",32);		# mm1 = carry5
+		&paddq("mm1","mm6");		# mm1 = carry5 + r[6] + w*a[6]
+		&movd(&DWP(24,$r,"",0),"mm1");
+		&psrlq("mm1",32);		# mm1 = carry6
+		&paddq("mm1","mm3");		# mm1 = carry6 + r[7] + w*a[7]
+		&movd(&DWP(28,$r,"",0),"mm1");
+		&add($r,32);
+		&psrlq("mm1",32);		# mm1 = carry_out
+
+		&sub("ecx",8);
+		&jnz(&label("maw_sse2_loop"));
+
+		&movd($c,"mm1");		# c = carry_out
+		&emms();
+
+		&jmp(&label("maw_finish"));
+	}
+
 	&set_label("maw_loop",0);

 	&mov(&swtmp(0),"ecx");	#
 <at>  <at>  -50,14 +126,14  <at>  <at> 
 		{
 		&comment("Round $i");

-		 &mov("eax",&DWP($i,$a,"",0)); 	# *a
+		&mov("eax",&DWP($i,$a,"",0)); 	# *a
 		&mul($w);			# *a * w
 		&add("eax",$c);		# L(t)+= *r
-		 &mov($c,&DWP($i,$r,"",0));	# L(t)+= *r
+		&mov($c,&DWP($i,$r,"",0));	# L(t)+= *r
 		&adc("edx",0);			# H(t)+=carry
-		 &add("eax",$c);		# L(t)+=c
+		&add("eax",$c);		# L(t)+=c
 		&adc("edx",0);			# H(t)+=carry
-		 &mov(&DWP($i,$r,"",0),"eax");	# *r= L(t);
+		&mov(&DWP($i,$r,"",0),"eax");	# *r= L(t);
 		&mov($c,"edx");			# c=  H(t);
 		}

===================================================================
RCS file: /home/dean/openssl/Repository/openssl/crypto/perlasm/x86asm.pl,v
retrieving revision 1.10
diff -u -r1.10 x86asm.pl
--- crypto/perlasm/x86asm.pl.orig	28 Nov 2003 13:10:47 -0000	1.10
+++ crypto/perlasm/x86asm.pl	8 Dec 2003 04:26:27 -0000
 <at>  <at>  -54,7 +54,11  <at>  <at> 
 		}

 	$pic=0;
-	for ( <at> ARGV) {	$pic=1 if (/\-[fK]PIC/i);	}
+	$cpuid=0;
+	for ( <at> ARGV) {
+		$pic=1 if (/\-[fK]PIC/i);
+		$cpuid=1 if (/\-DCPUID/);
+	}

 	&asm_init_output();

===================================================================
RCS file: /home/dean/openssl/Repository/openssl/crypto/perlasm/x86ms.pl,v
retrieving revision 1.8
diff -u -r1.8 x86ms.pl
--- crypto/perlasm/x86ms.pl.orig	27 Sep 2003 22:14:39 -0000	1.8
+++ crypto/perlasm/x86ms.pl	8 Dec 2003 04:34:19 -0000
 <at>  <at>  -120,6 +120,7  <at>  <at> 
 sub main'mov	{ &out2("mov", <at> _); }
 sub main'movb	{ &out2("mov", <at> _); }
 sub main'and	{ &out2("and", <at> _); }
+sub main'test	{ &out2("test", <at> _); }
 sub main'or	{ &out2("or", <at> _); }
 sub main'shl	{ &out2("shl", <at> _); }
 sub main'shr	{ &out2("shr", <at> _); }
 <at>  <at>  -161,6 +162,13  <at>  <at> 
 sub main'ret	{ &out0("ret"); }
 sub main'nop	{ &out0("nop"); }

+sub main'emms	{ &out0("emms"); }
+sub main'movd	{ &out2("movd", <at> _); }
+sub main'paddq	{ &out2("paddq", <at> _); }
+sub main'pmuludq{ &out2("pmuludq", <at> _); }
+sub main'psrlq	{ &out2("psrlq", <at> _); }
+sub main'pxor	{ &out2("pxor", <at> _); }
+
 sub out2
 	{
 	local($name,$p1,$p2)= <at> _;
===================================================================
RCS file: /home/dean/openssl/Repository/openssl/crypto/perlasm/x86mwasm_nw.pl,v
retrieving revision 1.1
diff -u -r1.1 x86mwasm_nw.pl
--- crypto/perlasm/x86mwasm_nw.pl.orig	28 Nov 2003 13:10:47 -0000	1.1
+++ crypto/perlasm/x86mwasm_nw.pl	8 Dec 2003 04:34:54 -0000
 <at>  <at>  -134,6 +134,7  <at>  <at> 
 sub main'mov	{ &out2("mov", <at> _); }
 sub main'movb	{ &out2("mov", <at> _); }
 sub main'and	{ &out2("and", <at> _); }
+sub main'test	{ &out2("test", <at> _); }
 sub main'or	{ &out2("or", <at> _); }
 sub main'shl	{ &out2("shl", <at> _); }
 sub main'shr	{ &out2("shr", <at> _); }
 <at>  <at>  -177,6 +178,13  <at>  <at> 
 sub main'ret	{ &out0("ret"); }
 sub main'nop	{ &out0("nop"); }

+sub main'emms	{ &out0("emms"); }
+sub main'movd	{ &out2("movd", <at> _); }
+sub main'paddq	{ &out2("paddq", <at> _); }
+sub main'pmuludq{ &out2("pmuludq", <at> _); }
+sub main'psrlq	{ &out2("psrlq", <at> _); }
+sub main'pxor	{ &out2("pxor", <at> _); }
+
 sub out2
 	{
 	my($name,$p1,$p2)= <at> _;
===================================================================
RCS file: /home/dean/openssl/Repository/openssl/crypto/perlasm/x86nasm.pl,v
retrieving revision 1.8
diff -u -r1.8 x86nasm.pl
--- crypto/perlasm/x86nasm.pl.orig	27 Sep 2003 22:14:39 -0000	1.8
+++ crypto/perlasm/x86nasm.pl	8 Dec 2003 04:34:02 -0000
 <at>  <at>  -126,6 +126,7  <at>  <at> 
 sub main'mov	{ &out2("mov", <at> _); }
 sub main'movb	{ &out2("mov", <at> _); }
 sub main'and	{ &out2("and", <at> _); }
+sub main'test	{ &out2("test", <at> _); }
 sub main'or	{ &out2("or", <at> _); }
 sub main'shl	{ &out2("shl", <at> _); }
 sub main'shr	{ &out2("shr", <at> _); }
 <at>  <at>  -170,6 +171,13  <at>  <at> 
 sub main'ret	{ &out0("ret"); }
 sub main'nop	{ &out0("nop"); }

+sub main'emms	{ &out0("emms"); }
+sub main'movd	{ &out2("movd", <at> _); }
+sub main'paddq	{ &out2("paddq", <at> _); }
+sub main'pmuludq{ &out2("pmuludq", <at> _); }
+sub main'psrlq	{ &out2("psrlq", <at> _); }
+sub main'pxor	{ &out2("pxor", <at> _); }
+
 sub out2
 	{
 	my($name,$p1,$p2)= <at> _;
===================================================================
RCS file: /home/dean/openssl/Repository/openssl/crypto/perlasm/x86unix.pl,v
retrieving revision 1.23
diff -u -r1.23 x86unix.pl
--- crypto/perlasm/x86unix.pl.orig	27 Sep 2003 22:14:39 -0000	1.23
+++ crypto/perlasm/x86unix.pl	8 Dec 2003 04:33:48 -0000
 <at>  <at>  -51,6 +51,14  <at>  <at> 
 	'edi',	'%edi',
 	'ebp',	'%ebp',
 	'esp',	'%esp',
+	'mm0',	'%mm0',
+	'mm1',	'%mm1',
+	'mm2',	'%mm2',
+	'mm3',	'%mm3',
+	'mm4',	'%mm4',
+	'mm5',	'%mm5',
+	'mm6',	'%mm6',
+	'mm7',	'%mm7',
 	);

 %reg_val=(
 <at>  <at>  -132,6 +140,7  <at>  <at> 
 sub main'mov	{ &out2("movl", <at> _); }
 sub main'movb	{ &out2("movb", <at> _); }
 sub main'and	{ &out2("andl", <at> _); }
+sub main'test	{ &out2("testl", <at> _); }
 sub main'or	{ &out2("orl", <at> _); }
 sub main'shl	{ &out2("sall", <at> _); }
 sub main'shr	{ &out2("shrl", <at> _); }
 <at>  <at>  -174,6 +183,13  <at>  <at> 
 sub main'ret	{ &out0("ret"); }
 sub main'nop	{ &out0("nop"); }

+sub main'emms	{ &out0("emms"); }
+sub main'movd	{ &out2("movd", <at> _); }
+sub main'paddq	{ &out2("paddq", <at> _); }
+sub main'pmuludq{ &out2("pmuludq", <at> _); }
+sub main'psrlq	{ &out2("psrlq", <at> _); }
+sub main'pxor	{ &out2("pxor", <at> _); }
+
 # The bswapl instruction is new for the 486. Emulate if i386.
 sub main'bswap
 	{
 <at>  <at>  -586,3 +602,8  <at>  <at> 
 	}

 sub main'blindpop { &out1("popl", <at> _); }
+
+$main'cpuid_mmx = 2;
+$main'cpuid_sse = 4;
+$main'cpuid_sse2 = 8;
+$main'cpuid_sse3 = 16;

______________________________________________________________________
OpenSSL Project                                 http://www.openssl.org
Development Mailing List                       openssl-dev <at> openssl.org
Automated List Manager                           majordomo <at> openssl.org

[openssl.org #838] REQ: Creating a BIO from a FILE* should retain (TEXT) access modes

Type: REQ
OS: WinNT
Version: 0.97c

Brief: A text mode FILE* will result in a binary BIO with BIO_set_fp.

Descr: This seems to make sense since BIO_set_fp also takes in text mode
flags along with the FILE*, but on the PEM_write_xx FILE* fns, there is no
way to set the text flag when passing in the FILE*. This has led to garbage
(bad newlines) certificates being generated under Win2K with a custom
certificate generation utility implemented with OpenSSL.

If the generated certs aren't used by a program that tries UNIX and MSDOS
newlines, this basically forces the developer to use the BIO routines
instead, making the write routines with FILE* useless.

______________________________________________________________________
OpenSSL Project                                 http://www.openssl.org
Development Mailing List                       openssl-dev <at> openssl.org
Automated List Manager                           majordomo <at> openssl.org

Re: [openssl.org #838] REQ: Creating a BIO from a FILE* should retain (TEXT) access modes

This one is related

tohttp://www.mail-archive.com/openssl-dev <at> openssl.org/msg16850.html

Steve.Chan <at> sybase.com via RT wrote:
> Type: REQ
> OS: WinNT
> Version: 0.97c
> 
> Brief: A text mode FILE* will result in a binary BIO with BIO_set_fp.
> 
> Descr: This seems to make sense since BIO_set_fp also takes in text mode
> flags along with the FILE*, but on the PEM_write_xx FILE* fns, there is no
> way to set the text flag when passing in the FILE*. This has led to garbage
> (bad newlines) certificates being generated under Win2K with a custom
> certificate generation utility implemented with OpenSSL.
> 
> If the generated certs aren't used by a program that tries UNIX and MSDOS
> newlines, this basically forces the developer to use the BIO routines
> instead, making the write routines with FILE* useless.
> 
> 
> ______________________________________________________________________
> OpenSSL Project                                 http://www.openssl.org
> Development Mailing List                       openssl-dev <at> openssl.org
> Automated List Manager                           majordomo <at> openssl.org

--

-- 
Lev Walkin
vlm <at> netli.com
______________________________________________________________________
OpenSSL Project                                 http://www.openssl.org
Development Mailing List                       openssl-dev <at> openssl.org
Automated List Manager                           majordomo <at> openssl.org

[openssl.org #839] Patch pk7_lib.c:PKCS7_get_recip_info()

______________________________________________________________________
OpenSSL Project                                 http://www.openssl.org
Development Mailing List                       openssl-dev <at> openssl.org
Automated List Manager                           majordomo <at> openssl.org