[PATCH][dizzy] nss: define MP_USE_UINT_DIGIT when compile mpi_x86.s

[PATCH][dizzy] nss: define MP_USE_UINT_DIGIT when compile mpi_x86.s

[Mark Hatle]

From: Roy Li <rongqing.li@windriver.com>

Backport patch to fix the segfault when do multiple.

Signed-off-by: Roy Li <rongqing.li@windriver.com>
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
---
 .../nss/files/Using-correct-macro-for-x86.patch    | 42 ++++++++++++++++++++++
 meta/recipes-support/nss/nss.inc                   |  1 +
 2 files changed, 43 insertions(+)
 create mode 100644 meta/recipes-support/nss/files/Using-correct-macro-for-x86.patch

diff --git a/meta/recipes-support/nss/files/Using-correct-macro-for-x86.patch b/meta/recipes-support/nss/files/Using-correct-macro-for-x86.patch
new file mode 100644
index 0000000..21d2ef7
--- /dev/null
+++ b/meta/recipes-support/nss/files/Using-correct-macro-for-x86.patch
@@ -0,0 +1,42 @@
+Using correct macro for x86
+
+Upstream-Status: Backport
+
+squash 11040 and 11043 as one patch
+
+changeset:   11040:60412aa39b8d
+user:        Stephan Bergmann <sbergman@redhat.com>
+date:        Fri Feb 14 17:45:40 2014 -0800
+files:       lib/freebl/Makefile
+description:
+Bug 972450: Define -DMP_USE_UINT_DIGIT for Linux x86 because mpi_x86.s
+requires mp_digit to be an unsigned int. The MP_USE_UINT_DIGIT macro
+ensures that. We were previously relying on ULLONG_MAX being undefined
+in the compilation environment, which isn't true in C99 mode. Also
+define -DMP_NO_MP_WORD to match Windows x86. r=wtc.
+
+
+changeset:   11043:fa9a3a970910
+parent:      11040:60412aa39b8d
+user:        Wan-Teh Chang <wtc@google.com>
+date:        Tue Feb 18 14:26:25 2014 -0800
+files:       lib/freebl/Makefile
+description:
+Bug 972450: Don't define -DMP_NO_MP_WORD for Linux x86 (unless one
+can show it improves performance).
+
+Signed-off-by: Roy Li <rongqing.li@windriver.com>
+
+diff -r f451ea1a2b12 -r 60412aa39b8d nss/lib/freebl/Makefile
+--- a/nss/lib/freebl/Makefile	Thu Feb 13 18:55:13 2014 -0800
++++ b/nss/lib/freebl/Makefile	Fri Feb 14 17:45:40 2014 -0800
+@@ -195,7 +195,7 @@
+ ifeq ($(CPU_ARCH),x86)
+     ASFILES  = mpi_x86.s
+     DEFINES += -DMP_ASSEMBLY_MULTIPLY -DMP_ASSEMBLY_SQUARE 
+-    DEFINES += -DMP_ASSEMBLY_DIV_2DX1D
++    DEFINES += -DMP_ASSEMBLY_DIV_2DX1D -DMP_USE_UINT_DIGIT
+     DEFINES += -DMP_CHAR_STORE_SLOW -DMP_IS_LITTLE_ENDIAN
+     # The floating point ECC code doesn't work on Linux x86 (bug 311432).
+     #ECL_USE_FP = 1
+
diff --git a/meta/recipes-support/nss/nss.inc b/meta/recipes-support/nss/nss.inc
index 7919ddd..c284e63 100644
--- a/meta/recipes-support/nss/nss.inc
+++ b/meta/recipes-support/nss/nss.inc
@@ -24,6 +24,7 @@ SRC_URI = "\
     file://nss-CVE-2013-5606.patch \
     file://nss-CVE-2014-1544.patch \
     file://nss-CVE-2014-1568.patch \
+    file://Using-correct-macro-for-x86.patch \
 "
 SRC_URI_append = "\
     file://nss.pc.in \
-- 
1.9.3

Re: [PATCH][dizzy] nss: define MP_USE_UINT_DIGIT when compile mpi_x86.s

[akuster808]

thanks,

staging for dizzy merge.

- armin

On 11/18/2014 06:41 PM, Mark Hatle wrote:
> From: Roy Li <rongqing.li@windriver.com>
>
> Backport patch to fix the segfault when do multiple.
>
> Signed-off-by: Roy Li <rongqing.li@windriver.com>
> Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
> ---
>   .../nss/files/Using-correct-macro-for-x86.patch    | 42 ++++++++++++++++++++++
>   meta/recipes-support/nss/nss.inc                   |  1 +
>   2 files changed, 43 insertions(+)
>   create mode 100644 meta/recipes-support/nss/files/Using-correct-macro-for-x86.patch
>
> diff --git a/meta/recipes-support/nss/files/Using-correct-macro-for-x86.patch b/meta/recipes-support/nss/files/Using-correct-macro-for-x86.patch
> new file mode 100644
> index 0000000..21d2ef7
> --- /dev/null
> +++ b/meta/recipes-support/nss/files/Using-correct-macro-for-x86.patch
> @@ -0,0 +1,42 @@
> +Using correct macro for x86
> +
> +Upstream-Status: Backport
> +
> +squash 11040 and 11043 as one patch
> +
> +changeset:   11040:60412aa39b8d
> +user:        Stephan Bergmann <sbergman@redhat.com>
> +date:        Fri Feb 14 17:45:40 2014 -0800
> +files:       lib/freebl/Makefile
> +description:
> +Bug 972450: Define -DMP_USE_UINT_DIGIT for Linux x86 because mpi_x86.s
> +requires mp_digit to be an unsigned int. The MP_USE_UINT_DIGIT macro
> +ensures that. We were previously relying on ULLONG_MAX being undefined
> +in the compilation environment, which isn't true in C99 mode. Also
> +define -DMP_NO_MP_WORD to match Windows x86. r=wtc.
> +
> +
> +changeset:   11043:fa9a3a970910
> +parent:      11040:60412aa39b8d
> +user:        Wan-Teh Chang <wtc@google.com>
> +date:        Tue Feb 18 14:26:25 2014 -0800
> +files:       lib/freebl/Makefile
> +description:
> +Bug 972450: Don't define -DMP_NO_MP_WORD for Linux x86 (unless one
> +can show it improves performance).
> +
> +Signed-off-by: Roy Li <rongqing.li@windriver.com>
> +
> +diff -r f451ea1a2b12 -r 60412aa39b8d nss/lib/freebl/Makefile
> +--- a/nss/lib/freebl/Makefile	Thu Feb 13 18:55:13 2014 -0800
> ++++ b/nss/lib/freebl/Makefile	Fri Feb 14 17:45:40 2014 -0800
> +@@ -195,7 +195,7 @@
> + ifeq ($(CPU_ARCH),x86)
> +     ASFILES  = mpi_x86.s
> +     DEFINES += -DMP_ASSEMBLY_MULTIPLY -DMP_ASSEMBLY_SQUARE
> +-    DEFINES += -DMP_ASSEMBLY_DIV_2DX1D
> ++    DEFINES += -DMP_ASSEMBLY_DIV_2DX1D -DMP_USE_UINT_DIGIT
> +     DEFINES += -DMP_CHAR_STORE_SLOW -DMP_IS_LITTLE_ENDIAN
> +     # The floating point ECC code doesn't work on Linux x86 (bug 311432).
> +     #ECL_USE_FP = 1
> +
> diff --git a/meta/recipes-support/nss/nss.inc b/meta/recipes-support/nss/nss.inc
> index 7919ddd..c284e63 100644
> --- a/meta/recipes-support/nss/nss.inc
> +++ b/meta/recipes-support/nss/nss.inc
> @@ -24,6 +24,7 @@ SRC_URI = "\
>       file://nss-CVE-2013-5606.patch \
>       file://nss-CVE-2014-1544.patch \
>       file://nss-CVE-2014-1568.patch \
> +    file://Using-correct-macro-for-x86.patch \
>   "
>   SRC_URI_append = "\
>       file://nss.pc.in \
>