vpaes: LoongArch: Use getauxval(AT_HWCAP) for LSX detection

Running LSX instructions requires both the hardware support and the
kernel support.  The `cpucfg` instruction only tests the hardware
support, causing a SIGILL if the hardware supports LSX but the kernel
does not.

Use `getauxval(AT_HWCAP)` as the ["Software Development and Build
Convention for LoongArch Architectures"][1] manual suggests.

The LOONGARCH_HWCAP_LSX and LOONGARCH_HWCAP_LASX bits are copied from
the manual too.  In Glibc 2.38 they'll be provided by <sys/auxv.h> as
well, but they are unavailable in earlier Glibc versions so we cannot
rely on it.

The getauxval syscall and Glibc wrapper are available since day one
(Linux-5.19 and Glibc-2.36) for LoongArch.

Fixes #21508.

[1]:https://github.com/loongson/la-softdev-convention/blob/master/la-softdev-convention.adoc#kernel-constraints

Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Tom Cosgrove <tom.cosgrove@arm.com>
Reviewed-by: Paul Dale <pauli@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/21509)

diff --git a/crypto/loongarch_arch.h b/crypto/loongarch_arch.h
index 53caf53ef3e5b3d279c8f043b977a631837f78f8..454be913329cbb3032516eacd800eb401e220081 100644 (file)
--- a/crypto/loongarch_arch.h
+++ b/crypto/loongarch_arch.h
@@ -9,9 +9,8 @@
 #ifndef OSSL_CRYPTO_LOONGARCH_ARCH_H
 # define OSSL_CRYPTO_LOONGARCH_ARCH_H
 
-extern unsigned int OPENSSL_loongarchcap_P;
-# define LOONGARCH_CFG2      0x02
-# define LOONGARCH_CFG2_LSX  (1<<6)
-# define LOONGARCH_CFG2_LASX (1<<7)
+extern unsigned int OPENSSL_loongarch_hwcap_P;
+# define LOONGARCH_HWCAP_LSX  (1 << 4)
+# define LOONGARCH_HWCAP_LASX (1 << 5)
 
 #endif

diff --git a/crypto/loongarchcap.c b/crypto/loongarchcap.c
index 67e3c02b41ab1f91ad6dfb49fb70d06aa989fca9..89839099386be61e85605305f8160b8bcf037f2f 100644 (file)
--- a/crypto/loongarchcap.c
+++ b/crypto/loongarchcap.c
@@ -6,17 +6,12 @@
  * in the file LICENSE in the source distribution or at
  * https://www.openssl.org/source/license.html
  */
+#include <sys/auxv.h>
 #include "loongarch_arch.h"
 
-unsigned int OPENSSL_loongarchcap_P = 0;
+unsigned int OPENSSL_loongarch_hwcap_P = 0;
 
 void OPENSSL_cpuid_setup(void)
 {
-       unsigned int reg;
-       __asm__ volatile(
-           "cpucfg %0, %1 \n\t"
-           : "+&r"(reg)
-           : "r"(LOONGARCH_CFG2)
-       );
-       OPENSSL_loongarchcap_P = reg;
+       OPENSSL_loongarch_hwcap_P = getauxval(AT_HWCAP);
 }

diff --git a/include/crypto/aes_platform.h b/include/crypto/aes_platform.h
index 2a6e4d3ec6290dc90eb79bbc66e7d92fdf0735a4..17cd9e3238345724fbbbd501584138a665bc7384 100644 (file)
--- a/include/crypto/aes_platform.h
+++ b/include/crypto/aes_platform.h
@@ -165,7 +165,7 @@ void gcm_ghash_v8(u64 Xi[2],const u128 Htable[16],const u8 *inp, size_t len);
 # if defined(__loongarch__) || defined(__loongarch64)
 #  include "loongarch_arch.h"
 #  if defined(VPAES_ASM)
-#   define VPAES_CAPABLE  (OPENSSL_loongarchcap_P & LOONGARCH_CFG2_LSX)
+#   define VPAES_CAPABLE  (OPENSSL_loongarch_hwcap_P & LOONGARCH_HWCAP_LSX)
 #  endif
 # endif