Merge tag 'net-next-6.3' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-nextgrafted

Pull networking updates from Jakub Kicinski:
 "Core:

   - Add dedicated kmem_cache for typical/small skb->head, avoid having
     to access struct page at kfree time, and improve memory use.

   - Introduce sysctl to set default RPS configuration for new netdevs.

   - Define Netlink protocol specification format which can be used to
     describe messages used by each family and auto-generate parsers.
     Add tools for generating kernel data structures and uAPI headers.

   - Expose all net/core sysctls inside netns.

   - Remove 4s sleep in netpoll if carrier is instantly detected on
     boot.

   - Add configurable limit of MDB entries per port, and port-vlan.

   - Continue populating drop reasons throughout the stack.

   - Retire a handful of legacy Qdiscs and classifiers.

  Protocols:

   - Support IPv4 big TCP (TSO frames larger than 64kB).

   - Add IP_LOCAL_PORT_RANGE socket option, to control local port range
     on socket by socket basis.

   - Track and report in procfs number of MPTCP sockets used.

   - Support mixing IPv4 and IPv6 flows in the in-kernel MPTCP path
     manager.

   - IPv6: don't check net.ipv6.route.max_size and rely on garbage
     collection to free memory (similarly to IPv4).

   - Support Penultimate Segment Pop (PSP) flavor in SRv6 (RFC8986).

   - ICMP: add per-rate limit counters.

   - Add support for user scanning requests in ieee802154.

   - Remove static WEP support.

   - Support minimal Wi-Fi 7 Extremely High Throughput (EHT) rate
     reporting.

   - WiFi 7 EHT channel puncturing support (client & AP).

  BPF:

   - Add a rbtree data structure following the "next-gen data structure"
     precedent set by recently added linked list, that is, by using
     kfunc + kptr instead of adding a new BPF map type.

   - Expose XDP hints via kfuncs with initial support for RX hash and
     timestamp metadata.

   - Add BPF_F_NO_TUNNEL_KEY extension to bpf_skb_set_tunnel_key to
     better support decap on GRE tunnel devices not operating in collect
     metadata.

   - Improve x86 JIT's codegen for PROBE_MEM runtime error checks.

   - Remove the need for trace_printk_lock for bpf_trace_printk and
     bpf_trace_vprintk helpers.

   - Extend libbpf's bpf_tracing.h support for tracing arguments of
     kprobes/uprobes and syscall as a special case.

   - Significantly reduce the search time for module symbols by
     livepatch and BPF.

   - Enable cpumasks to be used as kptrs, which is useful for tracing
     programs tracking which tasks end up running on which CPUs in
     different time intervals.

   - Add support for BPF trampoline on s390x and riscv64.

   - Add capability to export the XDP features supported by the NIC.

   - Add __bpf_kfunc tag for marking kernel functions as kfuncs.

   - Add cgroup.memory=nobpf kernel parameter option to disable BPF
     memory accounting for container environments.

  Netfilter:

   - Remove the CLUSTERIP target. It has been marked as obsolete for
     years, and we still have WARN splats wrt races of the out-of-band
     /proc interface installed by this target.

   - Add 'destroy' commands to nf_tables. They are identical to the
     existing 'delete' commands, but do not return an error if the
     referenced object (set, chain, rule...) did not exist.

  Driver API:

   - Improve cpumask_local_spread() locality to help NICs set the right
     IRQ affinity on AMD platforms.

   - Separate C22 and C45 MDIO bus transactions more clearly.

   - Introduce new DCB table to control DSCP rewrite on egress.

   - Support configuration of Physical Layer Collision Avoidance (PLCA)
     Reconciliation Sublayer (RS) (802.3cg-2019). Modern version of
     shared medium Ethernet.

   - Support for MAC Merge layer (IEEE 802.3-2018 clause 99). Allowing
     preemption of low priority frames by high priority frames.

   - Add support for controlling MACSec offload using netlink SET.

   - Rework devlink instance refcounts to allow registration and
     de-registration under the instance lock. Split the code into
     multiple files, drop some of the unnecessarily granular locks and
     factor out common parts of netlink operation handling.

   - Add TX frame aggregation parameters (for USB drivers).

   - Add a new attr TCA_EXT_WARN_MSG to report TC (offload) warning
     messages with notifications for debug.

   - Allow offloading of UDP NEW connections via act_ct.

   - Add support for per action HW stats in TC.

   - Support hardware miss to TC action (continue processing in SW from
     a specific point in the action chain).

   - Warn if old Wireless Extension user space interface is used with
     modern cfg80211/mac80211 drivers. Do not support Wireless
     Extensions for Wi-Fi 7 devices at all. Everyone should switch to
     using nl80211 interface instead.

   - Improve the CAN bit timing configuration. Use extack to return
     error messages directly to user space, update the SJW handling,
     including the definition of a new default value that will benefit
     CAN-FD controllers, by increasing their oscillator tolerance.

  New hardware / drivers:

   - Ethernet:
      - nVidia BlueField-3 support (control traffic driver)
      - Ethernet support for imx93 SoCs
      - Motorcomm yt8531 gigabit Ethernet PHY
      - onsemi NCN26000 10BASE-T1S PHY (with support for PLCA)
      - Microchip LAN8841 PHY (incl. cable diagnostics and PTP)
      - Amlogic gxl MDIO mux

   - WiFi:
      - RealTek RTL8188EU (rtl8xxxu)
      - Qualcomm Wi-Fi 7 devices (ath12k)

   - CAN:
      - Renesas R-Car V4H

  Drivers:

   - Bluetooth:
      - Set Per Platform Antenna Gain (PPAG) for Intel controllers.

   - Ethernet NICs:
      - Intel (1G, igc):
         - support TSN / Qbv / packet scheduling features of i226 model
      - Intel (100G, ice):
         - use GNSS subsystem instead of TTY
         - multi-buffer XDP support
         - extend support for GPIO pins to E823 devices
      - nVidia/Mellanox:
         - update the shared buffer configuration on PFC commands
         - implement PTP adjphase function for HW offset control
         - TC support for Geneve and GRE with VF tunnel offload
         - more efficient crypto key management method
         - multi-port eswitch support
      - Netronome/Corigine:
         - add DCB IEEE support
         - support IPsec offloading for NFP3800
      - Freescale/NXP (enetc):
         - support XDP_REDIRECT for XDP non-linear buffers
         - improve reconfig, avoid link flap and waiting for idle
         - support MAC Merge layer
      - Other NICs:
         - sfc/ef100: add basic devlink support for ef100
         - ionic: rx_push mode operation (writing descriptors via MMIO)
         - bnxt: use the auxiliary bus abstraction for RDMA
         - r8169: disable ASPM and reset bus in case of tx timeout
         - cpsw: support QSGMII mode for J721e CPSW9G
         - cpts: support pulse-per-second output
         - ngbe: add an mdio bus driver
         - usbnet: optimize usbnet_bh() by avoiding unnecessary queuing
         - r8152: handle devices with FW with NCM support
         - amd-xgbe: support 10Mbps, 2.5GbE speeds and rx-adaptation
         - virtio-net: support multi buffer XDP
         - virtio/vsock: replace virtio_vsock_pkt with sk_buff
         - tsnep: XDP support

   - Ethernet high-speed switches:
      - nVidia/Mellanox (mlxsw):
         - add support for latency TLV (in FW control messages)
      - Microchip (sparx5):
         - separate explicit and implicit traffic forwarding rules, make
           the implicit rules always active
         - add support for egress DSCP rewrite
         - IS0 VCAP support (Ingress Classification)
         - IS2 VCAP filters (protos, L3 addrs, L4 ports, flags, ToS
           etc.)
         - ES2 VCAP support (Egress Access Control)
         - support for Per-Stream Filtering and Policing (802.1Q,
           8.6.5.1)

   - Ethernet embedded switches:
      - Marvell (mv88e6xxx):
         - add MAB (port auth) offload support
         - enable PTP receive for mv88e6390
      - NXP (ocelot):
         - support MAC Merge layer
         - support for the the vsc7512 internal copper phys
      - Microchip:
         - lan9303: convert to PHYLINK
         - lan966x: support TC flower filter statistics
         - lan937x: PTP support for KSZ9563/KSZ8563 and LAN937x
         - lan937x: support Credit Based Shaper configuration
         - ksz9477: support Energy Efficient Ethernet
      - other:
         - qca8k: convert to regmap read/write API, use bulk operations
         - rswitch: Improve TX timestamp accuracy

   - Intel WiFi (iwlwifi):
      - EHT (Wi-Fi 7) rate reporting
      - STEP equalizer support: transfer some STEP (connection to radio
        on platforms with integrated wifi) related parameters from the
        BIOS to the firmware.

   - Qualcomm 802.11ax WiFi (ath11k):
      - IPQ5018 support
      - Fine Timing Measurement (FTM) responder role support
      - channel 177 support

   - MediaTek WiFi (mt76):
      - per-PHY LED support
      - mt7996: EHT (Wi-Fi 7) support
      - Wireless Ethernet Dispatch (WED) reset support
      - switch to using page pool allocator

   - RealTek WiFi (rtw89):
      - support new version of Bluetooth co-existance

   - Mobile:
      - rmnet: support TX aggregation"

* tag 'net-next-6.3' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next: (1872 commits)
  page_pool: add a comment explaining the fragment counter usage
  net: ethtool: fix __ethtool_dev_mm_supported() implementation
  ethtool: pse-pd: Fix double word in comments
  xsk: add linux/vmalloc.h to xsk.c
  sefltests: netdevsim: wait for devlink instance after netns removal
  selftest: fib_tests: Always cleanup before exit
  net/mlx5e: Align IPsec ASO result memory to be as required by hardware
  net/mlx5e: TC, Set CT miss to the specific ct action instance
  net/mlx5e: Rename CHAIN_TO_REG to MAPPED_OBJ_TO_REG
  net/mlx5: Refactor tc miss handling to a single function
  net/mlx5: Kconfig: Make tc offload depend on tc skb extension
  net/sched: flower: Support hardware miss to tc action
  net/sched: flower: Move filter handle initialization earlier
  net/sched: cls_api: Support hardware miss to tc action
  net/sched: Rename user cookie and act cookie
  sfc: fix builds without CONFIG_RTC_LIB
  sfc: clean up some inconsistent indentings
  net/mlx4_en: Introduce flexible array to silence overflow warning
  net: lan966x: Fix possible deadlock inside PTP
  net/ulp: Remove redundant ->clone() test in inet_clone_ulp().
  ...

1 files changed, 1544 insertions, 0 deletions

diff --git a/arch/sparc/crypto/aes_asm.S b/arch/sparc/crypto/aes_asm.S
new file mode 100644
index 000000000..155cefb98
--- /dev/null
+++ b/arch/sparc/crypto/aes_asm.S
@@ -0,0 +1,1544 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#include <linux/linkage.h>
+#include <asm/visasm.h>
+
+#include "opcodes.h"
+
+#define ENCRYPT_TWO_ROUNDS(KEY_BASE, I0, I1, T0, T1) \
+	AES_EROUND01(KEY_BASE +  0, I0, I1, T0) \
+	AES_EROUND23(KEY_BASE +  2, I0, I1, T1) \
+	AES_EROUND01(KEY_BASE +  4, T0, T1, I0) \
+	AES_EROUND23(KEY_BASE +  6, T0, T1, I1)
+
+#define ENCRYPT_TWO_ROUNDS_2(KEY_BASE, I0, I1, I2, I3, T0, T1, T2, T3) \
+	AES_EROUND01(KEY_BASE +  0, I0, I1, T0) \
+	AES_EROUND23(KEY_BASE +  2, I0, I1, T1) \
+	AES_EROUND01(KEY_BASE +  0, I2, I3, T2) \
+	AES_EROUND23(KEY_BASE +  2, I2, I3, T3) \
+	AES_EROUND01(KEY_BASE +  4, T0, T1, I0) \
+	AES_EROUND23(KEY_BASE +  6, T0, T1, I1) \
+	AES_EROUND01(KEY_BASE +  4, T2, T3, I2) \
+	AES_EROUND23(KEY_BASE +  6, T2, T3, I3)
+
+#define ENCRYPT_TWO_ROUNDS_LAST(KEY_BASE, I0, I1, T0, T1) \
+	AES_EROUND01(KEY_BASE +  0, I0, I1, T0) \
+	AES_EROUND23(KEY_BASE +  2, I0, I1, T1) \
+	AES_EROUND01_L(KEY_BASE +  4, T0, T1, I0) \
+	AES_EROUND23_L(KEY_BASE +  6, T0, T1, I1)
+
+#define ENCRYPT_TWO_ROUNDS_LAST_2(KEY_BASE, I0, I1, I2, I3, T0, T1, T2, T3) \
+	AES_EROUND01(KEY_BASE +  0, I0, I1, T0) \
+	AES_EROUND23(KEY_BASE +  2, I0, I1, T1) \
+	AES_EROUND01(KEY_BASE +  0, I2, I3, T2) \
+	AES_EROUND23(KEY_BASE +  2, I2, I3, T3) \
+	AES_EROUND01_L(KEY_BASE +  4, T0, T1, I0) \
+	AES_EROUND23_L(KEY_BASE +  6, T0, T1, I1) \
+	AES_EROUND01_L(KEY_BASE +  4, T2, T3, I2) \
+	AES_EROUND23_L(KEY_BASE +  6, T2, T3, I3)
+
+	/* 10 rounds */
+#define ENCRYPT_128(KEY_BASE, I0, I1, T0, T1) \
+	ENCRYPT_TWO_ROUNDS(KEY_BASE +  0, I0, I1, T0, T1) \
+	ENCRYPT_TWO_ROUNDS(KEY_BASE +  8, I0, I1, T0, T1) \
+	ENCRYPT_TWO_ROUNDS(KEY_BASE + 16, I0, I1, T0, T1) \
+	ENCRYPT_TWO_ROUNDS(KEY_BASE + 24, I0, I1, T0, T1) \
+	ENCRYPT_TWO_ROUNDS_LAST(KEY_BASE + 32, I0, I1, T0, T1)
+
+#define ENCRYPT_128_2(KEY_BASE, I0, I1, I2, I3, T0, T1, T2, T3) \
+	ENCRYPT_TWO_ROUNDS_2(KEY_BASE +  0, I0, I1, I2, I3, T0, T1, T2, T3) \
+	ENCRYPT_TWO_ROUNDS_2(KEY_BASE +  8, I0, I1, I2, I3, T0, T1, T2, T3) \
+	ENCRYPT_TWO_ROUNDS_2(KEY_BASE + 16, I0, I1, I2, I3, T0, T1, T2, T3) \
+	ENCRYPT_TWO_ROUNDS_2(KEY_BASE + 24, I0, I1, I2, I3, T0, T1, T2, T3) \
+	ENCRYPT_TWO_ROUNDS_LAST_2(KEY_BASE + 32, I0, I1, I2, I3, T0, T1, T2, T3)
+
+	/* 12 rounds */
+#define ENCRYPT_192(KEY_BASE, I0, I1, T0, T1) \
+	ENCRYPT_TWO_ROUNDS(KEY_BASE +  0, I0, I1, T0, T1) \
+	ENCRYPT_TWO_ROUNDS(KEY_BASE +  8, I0, I1, T0, T1) \
+	ENCRYPT_TWO_ROUNDS(KEY_BASE + 16, I0, I1, T0, T1) \
+	ENCRYPT_TWO_ROUNDS(KEY_BASE + 24, I0, I1, T0, T1) \
+	ENCRYPT_TWO_ROUNDS(KEY_BASE + 32, I0, I1, T0, T1) \
+	ENCRYPT_TWO_ROUNDS_LAST(KEY_BASE + 40, I0, I1, T0, T1)
+
+#define ENCRYPT_192_2(KEY_BASE, I0, I1, I2, I3, T0, T1, T2, T3) \
+	ENCRYPT_TWO_ROUNDS_2(KEY_BASE +  0, I0, I1, I2, I3, T0, T1, T2, T3) \
+	ENCRYPT_TWO_ROUNDS_2(KEY_BASE +  8, I0, I1, I2, I3, T0, T1, T2, T3) \
+	ENCRYPT_TWO_ROUNDS_2(KEY_BASE + 16, I0, I1, I2, I3, T0, T1, T2, T3) \
+	ENCRYPT_TWO_ROUNDS_2(KEY_BASE + 24, I0, I1, I2, I3, T0, T1, T2, T3) \
+	ENCRYPT_TWO_ROUNDS_2(KEY_BASE + 32, I0, I1, I2, I3, T0, T1, T2, T3) \
+	ENCRYPT_TWO_ROUNDS_LAST_2(KEY_BASE + 40, I0, I1, I2, I3, T0, T1, T2, T3)
+
+	/* 14 rounds */
+#define ENCRYPT_256(KEY_BASE, I0, I1, T0, T1) \
+	ENCRYPT_TWO_ROUNDS(KEY_BASE +  0, I0, I1, T0, T1) \
+	ENCRYPT_TWO_ROUNDS(KEY_BASE +  8, I0, I1, T0, T1) \
+	ENCRYPT_TWO_ROUNDS(KEY_BASE + 16, I0, I1, T0, T1) \
+	ENCRYPT_TWO_ROUNDS(KEY_BASE + 24, I0, I1, T0, T1) \
+	ENCRYPT_TWO_ROUNDS(KEY_BASE + 32, I0, I1, T0, T1) \
+	ENCRYPT_TWO_ROUNDS(KEY_BASE + 40, I0, I1, T0, T1) \
+	ENCRYPT_TWO_ROUNDS_LAST(KEY_BASE + 48, I0, I1, T0, T1)
+
+#define ENCRYPT_256_TWO_ROUNDS_2(KEY_BASE, I0, I1, I2, I3, TMP_BASE) \
+	ENCRYPT_TWO_ROUNDS_2(KEY_BASE, I0, I1, I2, I3, \
+			     TMP_BASE + 0, TMP_BASE + 2, TMP_BASE + 4, TMP_BASE + 6)
+
+#define ENCRYPT_256_2(KEY_BASE, I0, I1, I2, I3) \
+	ENCRYPT_256_TWO_ROUNDS_2(KEY_BASE +  0, I0, I1, I2, I3, KEY_BASE + 48) \
+	ldd	[%o0 + 0xd0], %f56; \
+	ldd	[%o0 + 0xd8], %f58; \
+	ENCRYPT_256_TWO_ROUNDS_2(KEY_BASE +  8, I0, I1, I2, I3, KEY_BASE +  0) \
+	ldd	[%o0 + 0xe0], %f60; \
+	ldd	[%o0 + 0xe8], %f62; \
+	ENCRYPT_256_TWO_ROUNDS_2(KEY_BASE + 16, I0, I1, I2, I3, KEY_BASE +  0) \
+	ENCRYPT_256_TWO_ROUNDS_2(KEY_BASE + 24, I0, I1, I2, I3, KEY_BASE +  0) \
+	ENCRYPT_256_TWO_ROUNDS_2(KEY_BASE + 32, I0, I1, I2, I3, KEY_BASE +  0) \
+	ENCRYPT_256_TWO_ROUNDS_2(KEY_BASE + 40, I0, I1, I2, I3, KEY_BASE +  0) \
+	AES_EROUND01(KEY_BASE +  48, I0, I1, KEY_BASE + 0) \
+	AES_EROUND23(KEY_BASE +  50, I0, I1, KEY_BASE + 2) \
+	AES_EROUND01(KEY_BASE +  48, I2, I3, KEY_BASE + 4) \
+	AES_EROUND23(KEY_BASE +  50, I2, I3, KEY_BASE + 6) \
+	AES_EROUND01_L(KEY_BASE +  52, KEY_BASE + 0, KEY_BASE + 2, I0) \
+	AES_EROUND23_L(KEY_BASE +  54, KEY_BASE + 0, KEY_BASE + 2, I1) \
+	ldd	[%o0 + 0x10], %f8; \
+	ldd	[%o0 + 0x18], %f10; \
+	AES_EROUND01_L(KEY_BASE +  52, KEY_BASE + 4, KEY_BASE + 6, I2) \
+	AES_EROUND23_L(KEY_BASE +  54, KEY_BASE + 4, KEY_BASE + 6, I3) \
+	ldd	[%o0 + 0x20], %f12; \
+	ldd	[%o0 + 0x28], %f14;
+
+#define DECRYPT_TWO_ROUNDS(KEY_BASE, I0, I1, T0, T1) \
+	AES_DROUND23(KEY_BASE +  0, I0, I1, T1) \
+	AES_DROUND01(KEY_BASE +  2, I0, I1, T0) \
+	AES_DROUND23(KEY_BASE +  4, T0, T1, I1) \
+	AES_DROUND01(KEY_BASE +  6, T0, T1, I0)
+
+#define DECRYPT_TWO_ROUNDS_2(KEY_BASE, I0, I1, I2, I3, T0, T1, T2, T3) \
+	AES_DROUND23(KEY_BASE +  0, I0, I1, T1) \
+	AES_DROUND01(KEY_BASE +  2, I0, I1, T0) \
+	AES_DROUND23(KEY_BASE +  0, I2, I3, T3) \
+	AES_DROUND01(KEY_BASE +  2, I2, I3, T2) \
+	AES_DROUND23(KEY_BASE +  4, T0, T1, I1) \
+	AES_DROUND01(KEY_BASE +  6, T0, T1, I0) \
+	AES_DROUND23(KEY_BASE +  4, T2, T3, I3) \
+	AES_DROUND01(KEY_BASE +  6, T2, T3, I2)
+
+#define DECRYPT_TWO_ROUNDS_LAST(KEY_BASE, I0, I1, T0, T1) \
+	AES_DROUND23(KEY_BASE +  0, I0, I1, T1) \
+	AES_DROUND01(KEY_BASE +  2, I0, I1, T0) \
+	AES_DROUND23_L(KEY_BASE +  4, T0, T1, I1) \
+	AES_DROUND01_L(KEY_BASE +  6, T0, T1, I0)
+
+#define DECRYPT_TWO_ROUNDS_LAST_2(KEY_BASE, I0, I1, I2, I3, T0, T1, T2, T3) \
+	AES_DROUND23(KEY_BASE +  0, I0, I1, T1) \
+	AES_DROUND01(KEY_BASE +  2, I0, I1, T0) \
+	AES_DROUND23(KEY_BASE +  0, I2, I3, T3) \
+	AES_DROUND01(KEY_BASE +  2, I2, I3, T2) \
+	AES_DROUND23_L(KEY_BASE +  4, T0, T1, I1) \
+	AES_DROUND01_L(KEY_BASE +  6, T0, T1, I0) \
+	AES_DROUND23_L(KEY_BASE +  4, T2, T3, I3) \
+	AES_DROUND01_L(KEY_BASE +  6, T2, T3, I2)
+
+	/* 10 rounds */
+#define DECRYPT_128(KEY_BASE, I0, I1, T0, T1) \
+	DECRYPT_TWO_ROUNDS(KEY_BASE +  0, I0, I1, T0, T1) \
+	DECRYPT_TWO_ROUNDS(KEY_BASE +  8, I0, I1, T0, T1) \
+	DECRYPT_TWO_ROUNDS(KEY_BASE + 16, I0, I1, T0, T1) \
+	DECRYPT_TWO_ROUNDS(KEY_BASE + 24, I0, I1, T0, T1) \
+	DECRYPT_TWO_ROUNDS_LAST(KEY_BASE + 32, I0, I1, T0, T1)
+
+#define DECRYPT_128_2(KEY_BASE, I0, I1, I2, I3, T0, T1, T2, T3) \
+	DECRYPT_TWO_ROUNDS_2(KEY_BASE +  0, I0, I1, I2, I3, T0, T1, T2, T3) \
+	DECRYPT_TWO_ROUNDS_2(KEY_BASE +  8, I0, I1, I2, I3, T0, T1, T2, T3) \
+	DECRYPT_TWO_ROUNDS_2(KEY_BASE + 16, I0, I1, I2, I3, T0, T1, T2, T3) \
+	DECRYPT_TWO_ROUNDS_2(KEY_BASE + 24, I0, I1, I2, I3, T0, T1, T2, T3) \
+	DECRYPT_TWO_ROUNDS_LAST_2(KEY_BASE + 32, I0, I1, I2, I3, T0, T1, T2, T3)
+
+	/* 12 rounds */
+#define DECRYPT_192(KEY_BASE, I0, I1, T0, T1) \
+	DECRYPT_TWO_ROUNDS(KEY_BASE +  0, I0, I1, T0, T1) \
+	DECRYPT_TWO_ROUNDS(KEY_BASE +  8, I0, I1, T0, T1) \
+	DECRYPT_TWO_ROUNDS(KEY_BASE + 16, I0, I1, T0, T1) \
+	DECRYPT_TWO_ROUNDS(KEY_BASE + 24, I0, I1, T0, T1) \
+	DECRYPT_TWO_ROUNDS(KEY_BASE + 32, I0, I1, T0, T1) \
+	DECRYPT_TWO_ROUNDS_LAST(KEY_BASE + 40, I0, I1, T0, T1)
+
+#define DECRYPT_192_2(KEY_BASE, I0, I1, I2, I3, T0, T1, T2, T3) \
+	DECRYPT_TWO_ROUNDS_2(KEY_BASE +  0, I0, I1, I2, I3, T0, T1, T2, T3) \
+	DECRYPT_TWO_ROUNDS_2(KEY_BASE +  8, I0, I1, I2, I3, T0, T1, T2, T3) \
+	DECRYPT_TWO_ROUNDS_2(KEY_BASE + 16, I0, I1, I2, I3, T0, T1, T2, T3) \
+	DECRYPT_TWO_ROUNDS_2(KEY_BASE + 24, I0, I1, I2, I3, T0, T1, T2, T3) \
+	DECRYPT_TWO_ROUNDS_2(KEY_BASE + 32, I0, I1, I2, I3, T0, T1, T2, T3) \
+	DECRYPT_TWO_ROUNDS_LAST_2(KEY_BASE + 40, I0, I1, I2, I3, T0, T1, T2, T3)
+
+	/* 14 rounds */
+#define DECRYPT_256(KEY_BASE, I0, I1, T0, T1) \
+	DECRYPT_TWO_ROUNDS(KEY_BASE +  0, I0, I1, T0, T1) \
+	DECRYPT_TWO_ROUNDS(KEY_BASE +  8, I0, I1, T0, T1) \
+	DECRYPT_TWO_ROUNDS(KEY_BASE + 16, I0, I1, T0, T1) \
+	DECRYPT_TWO_ROUNDS(KEY_BASE + 24, I0, I1, T0, T1) \
+	DECRYPT_TWO_ROUNDS(KEY_BASE + 32, I0, I1, T0, T1) \
+	DECRYPT_TWO_ROUNDS(KEY_BASE + 40, I0, I1, T0, T1) \
+	DECRYPT_TWO_ROUNDS_LAST(KEY_BASE + 48, I0, I1, T0, T1)
+
+#define DECRYPT_256_TWO_ROUNDS_2(KEY_BASE, I0, I1, I2, I3, TMP_BASE) \
+	DECRYPT_TWO_ROUNDS_2(KEY_BASE, I0, I1, I2, I3, \
+			     TMP_BASE + 0, TMP_BASE + 2, TMP_BASE + 4, TMP_BASE + 6)
+
+#define DECRYPT_256_2(KEY_BASE, I0, I1, I2, I3) \
+	DECRYPT_256_TWO_ROUNDS_2(KEY_BASE +  0, I0, I1, I2, I3, KEY_BASE + 48) \
+	ldd	[%o0 + 0x18], %f56; \
+	ldd	[%o0 + 0x10], %f58; \
+	DECRYPT_256_TWO_ROUNDS_2(KEY_BASE +  8, I0, I1, I2, I3, KEY_BASE +  0) \
+	ldd	[%o0 + 0x08], %f60; \
+	ldd	[%o0 + 0x00], %f62; \
+	DECRYPT_256_TWO_ROUNDS_2(KEY_BASE + 16, I0, I1, I2, I3, KEY_BASE +  0) \
+	DECRYPT_256_TWO_ROUNDS_2(KEY_BASE + 24, I0, I1, I2, I3, KEY_BASE +  0) \
+	DECRYPT_256_TWO_ROUNDS_2(KEY_BASE + 32, I0, I1, I2, I3, KEY_BASE +  0) \
+	DECRYPT_256_TWO_ROUNDS_2(KEY_BASE + 40, I0, I1, I2, I3, KEY_BASE +  0) \
+	AES_DROUND23(KEY_BASE +  48, I0, I1, KEY_BASE + 2) \
+	AES_DROUND01(KEY_BASE +  50, I0, I1, KEY_BASE + 0) \
+	AES_DROUND23(KEY_BASE +  48, I2, I3, KEY_BASE + 6) \
+	AES_DROUND01(KEY_BASE +  50, I2, I3, KEY_BASE + 4) \
+	AES_DROUND23_L(KEY_BASE +  52, KEY_BASE + 0, KEY_BASE + 2, I1) \
+	AES_DROUND01_L(KEY_BASE +  54, KEY_BASE + 0, KEY_BASE + 2, I0) \
+	ldd	[%o0 + 0xd8], %f8; \
+	ldd	[%o0 + 0xd0], %f10; \
+	AES_DROUND23_L(KEY_BASE +  52, KEY_BASE + 4, KEY_BASE + 6, I3) \
+	AES_DROUND01_L(KEY_BASE +  54, KEY_BASE + 4, KEY_BASE + 6, I2) \
+	ldd	[%o0 + 0xc8], %f12; \
+	ldd	[%o0 + 0xc0], %f14;
+
+	.align	32
+ENTRY(aes_sparc64_key_expand)
+	/* %o0=input_key, %o1=output_key, %o2=key_len */
+	VISEntry
+	ld	[%o0 + 0x00], %f0
+	ld	[%o0 + 0x04], %f1
+	ld	[%o0 + 0x08], %f2
+	ld	[%o0 + 0x0c], %f3
+
+	std	%f0, [%o1 + 0x00]
+	std	%f2, [%o1 + 0x08]
+	add	%o1, 0x10, %o1
+
+	cmp	%o2, 24
+	bl	2f
+	 nop
+
+	be	1f
+	 nop
+
+	/* 256-bit key expansion */
+	ld	[%o0 + 0x10], %f4
+	ld	[%o0 + 0x14], %f5
+	ld	[%o0 + 0x18], %f6
+	ld	[%o0 + 0x1c], %f7
+
+	std	%f4, [%o1 + 0x00]
+	std	%f6, [%o1 + 0x08]
+	add	%o1, 0x10, %o1
+
+	AES_KEXPAND1(0, 6, 0x0, 8)
+	AES_KEXPAND2(2, 8, 10)
+	AES_KEXPAND0(4, 10, 12)
+	AES_KEXPAND2(6, 12, 14)
+	AES_KEXPAND1(8, 14, 0x1, 16)
+	AES_KEXPAND2(10, 16, 18)
+	AES_KEXPAND0(12, 18, 20)
+	AES_KEXPAND2(14, 20, 22)
+	AES_KEXPAND1(16, 22, 0x2, 24)
+	AES_KEXPAND2(18, 24, 26)
+	AES_KEXPAND0(20, 26, 28)
+	AES_KEXPAND2(22, 28, 30)
+	AES_KEXPAND1(24, 30, 0x3, 32)
+	AES_KEXPAND2(26, 32, 34)
+	AES_KEXPAND0(28, 34, 36)
+	AES_KEXPAND2(30, 36, 38)
+	AES_KEXPAND1(32, 38, 0x4, 40)
+	AES_KEXPAND2(34, 40, 42)
+	AES_KEXPAND0(36, 42, 44)
+	AES_KEXPAND2(38, 44, 46)
+	AES_KEXPAND1(40, 46, 0x5, 48)
+	AES_KEXPAND2(42, 48, 50)
+	AES_KEXPAND0(44, 50, 52)
+	AES_KEXPAND2(46, 52, 54)
+	AES_KEXPAND1(48, 54, 0x6, 56)
+	AES_KEXPAND2(50, 56, 58)
+
+	std	%f8, [%o1 + 0x00]
+	std	%f10, [%o1 + 0x08]
+	std	%f12, [%o1 + 0x10]
+	std	%f14, [%o1 + 0x18]
+	std	%f16, [%o1 + 0x20]
+	std	%f18, [%o1 + 0x28]
+	std	%f20, [%o1 + 0x30]
+	std	%f22, [%o1 + 0x38]
+	std	%f24, [%o1 + 0x40]
+	std	%f26, [%o1 + 0x48]
+	std	%f28, [%o1 + 0x50]
+	std	%f30, [%o1 + 0x58]
+	std	%f32, [%o1 + 0x60]
+	std	%f34, [%o1 + 0x68]
+	std	%f36, [%o1 + 0x70]
+	std	%f38, [%o1 + 0x78]
+	std	%f40, [%o1 + 0x80]
+	std	%f42, [%o1 + 0x88]
+	std	%f44, [%o1 + 0x90]
+	std	%f46, [%o1 + 0x98]
+	std	%f48, [%o1 + 0xa0]
+	std	%f50, [%o1 + 0xa8]
+	std	%f52, [%o1 + 0xb0]
+	std	%f54, [%o1 + 0xb8]
+	std	%f56, [%o1 + 0xc0]
+	ba,pt	%xcc, 80f
+	 std	%f58, [%o1 + 0xc8]
+
+1:	
+	/* 192-bit key expansion */
+	ld	[%o0 + 0x10], %f4
+	ld	[%o0 + 0x14], %f5
+
+	std	%f4, [%o1 + 0x00]
+	add	%o1, 0x08, %o1
+
+	AES_KEXPAND1(0, 4, 0x0, 6)
+	AES_KEXPAND2(2, 6, 8)
+	AES_KEXPAND2(4, 8, 10)
+	AES_KEXPAND1(6, 10, 0x1, 12)
+	AES_KEXPAND2(8, 12, 14)
+	AES_KEXPAND2(10, 14, 16)
+	AES_KEXPAND1(12, 16, 0x2, 18)
+	AES_KEXPAND2(14, 18, 20)
+	AES_KEXPAND2(16, 20, 22)
+	AES_KEXPAND1(18, 22, 0x3, 24)
+	AES_KEXPAND2(20, 24, 26)
+	AES_KEXPAND2(22, 26, 28)
+	AES_KEXPAND1(24, 28, 0x4, 30)
+	AES_KEXPAND2(26, 30, 32)
+	AES_KEXPAND2(28, 32, 34)
+	AES_KEXPAND1(30, 34, 0x5, 36)
+	AES_KEXPAND2(32, 36, 38)
+	AES_KEXPAND2(34, 38, 40)
+	AES_KEXPAND1(36, 40, 0x6, 42)
+	AES_KEXPAND2(38, 42, 44)
+	AES_KEXPAND2(40, 44, 46)
+	AES_KEXPAND1(42, 46, 0x7, 48)
+	AES_KEXPAND2(44, 48, 50)
+
+	std	%f6, [%o1 + 0x00]
+	std	%f8, [%o1 + 0x08]
+	std	%f10, [%o1 + 0x10]
+	std	%f12, [%o1 + 0x18]
+	std	%f14, [%o1 + 0x20]
+	std	%f16, [%o1 + 0x28]
+	std	%f18, [%o1 + 0x30]
+	std	%f20, [%o1 + 0x38]
+	std	%f22, [%o1 + 0x40]
+	std	%f24, [%o1 + 0x48]
+	std	%f26, [%o1 + 0x50]
+	std	%f28, [%o1 + 0x58]
+	std	%f30, [%o1 + 0x60]
+	std	%f32, [%o1 + 0x68]
+	std	%f34, [%o1 + 0x70]
+	std	%f36, [%o1 + 0x78]
+	std	%f38, [%o1 + 0x80]
+	std	%f40, [%o1 + 0x88]
+	std	%f42, [%o1 + 0x90]
+	std	%f44, [%o1 + 0x98]
+	std	%f46, [%o1 + 0xa0]
+	std	%f48, [%o1 + 0xa8]
+	ba,pt	%xcc, 80f
+	 std	%f50, [%o1 + 0xb0]
+
+2:
+	/* 128-bit key expansion */
+	AES_KEXPAND1(0, 2, 0x0, 4)
+	AES_KEXPAND2(2, 4, 6)
+	AES_KEXPAND1(4, 6, 0x1, 8)
+	AES_KEXPAND2(6, 8, 10)
+	AES_KEXPAND1(8, 10, 0x2, 12)
+	AES_KEXPAND2(10, 12, 14)
+	AES_KEXPAND1(12, 14, 0x3, 16)
+	AES_KEXPAND2(14, 16, 18)
+	AES_KEXPAND1(16, 18, 0x4, 20)
+	AES_KEXPAND2(18, 20, 22)
+	AES_KEXPAND1(20, 22, 0x5, 24)
+	AES_KEXPAND2(22, 24, 26)
+	AES_KEXPAND1(24, 26, 0x6, 28)
+	AES_KEXPAND2(26, 28, 30)
+	AES_KEXPAND1(28, 30, 0x7, 32)
+	AES_KEXPAND2(30, 32, 34)
+	AES_KEXPAND1(32, 34, 0x8, 36)
+	AES_KEXPAND2(34, 36, 38)
+	AES_KEXPAND1(36, 38, 0x9, 40)
+	AES_KEXPAND2(38, 40, 42)
+
+	std	%f4, [%o1 + 0x00]
+	std	%f6, [%o1 + 0x08]
+	std	%f8, [%o1 + 0x10]
+	std	%f10, [%o1 + 0x18]
+	std	%f12, [%o1 + 0x20]
+	std	%f14, [%o1 + 0x28]
+	std	%f16, [%o1 + 0x30]
+	std	%f18, [%o1 + 0x38]
+	std	%f20, [%o1 + 0x40]
+	std	%f22, [%o1 + 0x48]
+	std	%f24, [%o1 + 0x50]
+	std	%f26, [%o1 + 0x58]
+	std	%f28, [%o1 + 0x60]
+	std	%f30, [%o1 + 0x68]
+	std	%f32, [%o1 + 0x70]
+	std	%f34, [%o1 + 0x78]
+	std	%f36, [%o1 + 0x80]
+	std	%f38, [%o1 + 0x88]
+	std	%f40, [%o1 + 0x90]
+	std	%f42, [%o1 + 0x98]
+80:
+	retl
+	 VISExit
+ENDPROC(aes_sparc64_key_expand)
+
+	.align		32
+ENTRY(aes_sparc64_encrypt_128)
+	/* %o0=key, %o1=input, %o2=output */
+	VISEntry
+	ld		[%o1 + 0x00], %f4
+	ld		[%o1 + 0x04], %f5
+	ld		[%o1 + 0x08], %f6
+	ld		[%o1 + 0x0c], %f7
+	ldd		[%o0 + 0x00], %f8
+	ldd		[%o0 + 0x08], %f10
+	ldd		[%o0 + 0x10], %f12
+	ldd		[%o0 + 0x18], %f14
+	ldd		[%o0 + 0x20], %f16
+	ldd		[%o0 + 0x28], %f18
+	ldd		[%o0 + 0x30], %f20
+	ldd		[%o0 + 0x38], %f22
+	ldd		[%o0 + 0x40], %f24
+	ldd		[%o0 + 0x48], %f26
+	ldd		[%o0 + 0x50], %f28
+	ldd		[%o0 + 0x58], %f30
+	ldd		[%o0 + 0x60], %f32
+	ldd		[%o0 + 0x68], %f34
+	ldd		[%o0 + 0x70], %f36
+	ldd		[%o0 + 0x78], %f38
+	ldd		[%o0 + 0x80], %f40
+	ldd		[%o0 + 0x88], %f42
+	ldd		[%o0 + 0x90], %f44
+	ldd		[%o0 + 0x98], %f46
+	ldd		[%o0 + 0xa0], %f48
+	ldd		[%o0 + 0xa8], %f50
+	fxor		%f8, %f4, %f4
+	fxor		%f10, %f6, %f6
+	ENCRYPT_128(12, 4, 6, 0, 2)
+	st		%f4, [%o2 + 0x00]
+	st		%f5, [%o2 + 0x04]
+	st		%f6, [%o2 + 0x08]
+	st		%f7, [%o2 + 0x0c]
+	retl
+	 VISExit
+ENDPROC(aes_sparc64_encrypt_128)
+
+	.align		32
+ENTRY(aes_sparc64_encrypt_192)
+	/* %o0=key, %o1=input, %o2=output */
+	VISEntry
+	ld		[%o1 + 0x00], %f4
+	ld		[%o1 + 0x04], %f5
+	ld		[%o1 + 0x08], %f6
+	ld		[%o1 + 0x0c], %f7
+
+	ldd		[%o0 + 0x00], %f8
+	ldd		[%o0 + 0x08], %f10
+
+	fxor		%f8, %f4, %f4
+	fxor		%f10, %f6, %f6
+
+	ldd		[%o0 + 0x10], %f8
+	ldd		[%o0 + 0x18], %f10
+	ldd		[%o0 + 0x20], %f12
+	ldd		[%o0 + 0x28], %f14
+	add		%o0, 0x20, %o0
+
+	ENCRYPT_TWO_ROUNDS(8, 4, 6, 0, 2)
+
+	ldd		[%o0 + 0x10], %f12
+	ldd		[%o0 + 0x18], %f14
+	ldd		[%o0 + 0x20], %f16
+	ldd		[%o0 + 0x28], %f18
+	ldd		[%o0 + 0x30], %f20
+	ldd		[%o0 + 0x38], %f22
+	ldd		[%o0 + 0x40], %f24
+	ldd		[%o0 + 0x48], %f26
+	ldd		[%o0 + 0x50], %f28
+	ldd		[%o0 + 0x58], %f30
+	ldd		[%o0 + 0x60], %f32
+	ldd		[%o0 + 0x68], %f34
+	ldd		[%o0 + 0x70], %f36
+	ldd		[%o0 + 0x78], %f38
+	ldd		[%o0 + 0x80], %f40
+	ldd		[%o0 + 0x88], %f42
+	ldd		[%o0 + 0x90], %f44
+	ldd		[%o0 + 0x98], %f46
+	ldd		[%o0 + 0xa0], %f48
+	ldd		[%o0 + 0xa8], %f50
+
+
+	ENCRYPT_128(12, 4, 6, 0, 2)
+
+	st		%f4, [%o2 + 0x00]
+	st		%f5, [%o2 + 0x04]
+	st		%f6, [%o2 + 0x08]
+	st		%f7, [%o2 + 0x0c]
+
+	retl
+	 VISExit
+ENDPROC(aes_sparc64_encrypt_192)
+
+	.align		32
+ENTRY(aes_sparc64_encrypt_256)
+	/* %o0=key, %o1=input, %o2=output */
+	VISEntry
+	ld		[%o1 + 0x00], %f4
+	ld		[%o1 + 0x04], %f5
+	ld		[%o1 + 0x08], %f6
+	ld		[%o1 + 0x0c], %f7
+
+	ldd		[%o0 + 0x00], %f8
+	ldd		[%o0 + 0x08], %f10
+
+	fxor		%f8, %f4, %f4
+	fxor		%f10, %f6, %f6
+
+	ldd		[%o0 + 0x10], %f8
+
+	ldd		[%o0 + 0x18], %f10
+	ldd		[%o0 + 0x20], %f12
+	ldd		[%o0 + 0x28], %f14
+	add		%o0, 0x20, %o0
+
+	ENCRYPT_TWO_ROUNDS(8, 4, 6, 0, 2)
+
+	ldd		[%o0 + 0x10], %f8
+
+	ldd		[%o0 + 0x18], %f10
+	ldd		[%o0 + 0x20], %f12
+	ldd		[%o0 + 0x28], %f14
+	add		%o0, 0x20, %o0
+
+	ENCRYPT_TWO_ROUNDS(8, 4, 6, 0, 2)
+
+	ldd		[%o0 + 0x10], %f12
+	ldd		[%o0 + 0x18], %f14
+	ldd		[%o0 + 0x20], %f16
+	ldd		[%o0 + 0x28], %f18
+	ldd		[%o0 + 0x30], %f20
+	ldd		[%o0 + 0x38], %f22
+	ldd		[%o0 + 0x40], %f24
+	ldd		[%o0 + 0x48], %f26
+	ldd		[%o0 + 0x50], %f28
+	ldd		[%o0 + 0x58], %f30
+	ldd		[%o0 + 0x60], %f32
+	ldd		[%o0 + 0x68], %f34
+	ldd		[%o0 + 0x70], %f36
+	ldd		[%o0 + 0x78], %f38
+	ldd		[%o0 + 0x80], %f40
+	ldd		[%o0 + 0x88], %f42
+	ldd		[%o0 + 0x90], %f44
+	ldd		[%o0 + 0x98], %f46
+	ldd		[%o0 + 0xa0], %f48
+	ldd		[%o0 + 0xa8], %f50
+
+	ENCRYPT_128(12, 4, 6, 0, 2)
+
+	st		%f4, [%o2 + 0x00]
+	st		%f5, [%o2 + 0x04]
+	st		%f6, [%o2 + 0x08]
+	st		%f7, [%o2 + 0x0c]
+
+	retl
+	 VISExit
+ENDPROC(aes_sparc64_encrypt_256)
+
+	.align		32
+ENTRY(aes_sparc64_decrypt_128)
+	/* %o0=key, %o1=input, %o2=output */
+	VISEntry
+	ld		[%o1 + 0x00], %f4
+	ld		[%o1 + 0x04], %f5
+	ld		[%o1 + 0x08], %f6
+	ld		[%o1 + 0x0c], %f7
+	ldd		[%o0 + 0xa0], %f8
+	ldd		[%o0 + 0xa8], %f10
+	ldd		[%o0 + 0x98], %f12
+	ldd		[%o0 + 0x90], %f14
+	ldd		[%o0 + 0x88], %f16
+	ldd		[%o0 + 0x80], %f18
+	ldd		[%o0 + 0x78], %f20
+	ldd		[%o0 + 0x70], %f22
+	ldd		[%o0 + 0x68], %f24
+	ldd		[%o0 + 0x60], %f26
+	ldd		[%o0 + 0x58], %f28
+	ldd		[%o0 + 0x50], %f30
+	ldd		[%o0 + 0x48], %f32
+	ldd		[%o0 + 0x40], %f34
+	ldd		[%o0 + 0x38], %f36
+	ldd		[%o0 + 0x30], %f38
+	ldd		[%o0 + 0x28], %f40
+	ldd		[%o0 + 0x20], %f42
+	ldd		[%o0 + 0x18], %f44
+	ldd		[%o0 + 0x10], %f46
+	ldd		[%o0 + 0x08], %f48
+	ldd		[%o0 + 0x00], %f50
+	fxor		%f8, %f4, %f4
+	fxor		%f10, %f6, %f6
+	DECRYPT_128(12, 4, 6, 0, 2)
+	st		%f4, [%o2 + 0x00]
+	st		%f5, [%o2 + 0x04]
+	st		%f6, [%o2 + 0x08]
+	st		%f7, [%o2 + 0x0c]
+	retl
+	 VISExit
+ENDPROC(aes_sparc64_decrypt_128)
+
+	.align		32
+ENTRY(aes_sparc64_decrypt_192)
+	/* %o0=key, %o1=input, %o2=output */
+	VISEntry
+	ld		[%o1 + 0x00], %f4
+	ld		[%o1 + 0x04], %f5
+	ld		[%o1 + 0x08], %f6
+	ld		[%o1 + 0x0c], %f7
+	ldd		[%o0 + 0xc0], %f8
+	ldd		[%o0 + 0xc8], %f10
+	ldd		[%o0 + 0xb8], %f12
+	ldd		[%o0 + 0xb0], %f14
+	ldd		[%o0 + 0xa8], %f16
+	ldd		[%o0 + 0xa0], %f18
+	fxor		%f8, %f4, %f4
+	fxor		%f10, %f6, %f6
+	ldd		[%o0 + 0x98], %f20
+	ldd		[%o0 + 0x90], %f22
+	ldd		[%o0 + 0x88], %f24
+	ldd		[%o0 + 0x80], %f26
+	DECRYPT_TWO_ROUNDS(12, 4, 6, 0, 2)
+	ldd		[%o0 + 0x78], %f28
+	ldd		[%o0 + 0x70], %f30
+	ldd		[%o0 + 0x68], %f32
+	ldd		[%o0 + 0x60], %f34
+	ldd		[%o0 + 0x58], %f36
+	ldd		[%o0 + 0x50], %f38
+	ldd		[%o0 + 0x48], %f40
+	ldd		[%o0 + 0x40], %f42
+	ldd		[%o0 + 0x38], %f44
+	ldd		[%o0 + 0x30], %f46
+	ldd		[%o0 + 0x28], %f48
+	ldd		[%o0 + 0x20], %f50
+	ldd		[%o0 + 0x18], %f52
+	ldd		[%o0 + 0x10], %f54
+	ldd		[%o0 + 0x08], %f56
+	ldd		[%o0 + 0x00], %f58
+	DECRYPT_128(20, 4, 6, 0, 2)
+	st		%f4, [%o2 + 0x00]
+	st		%f5, [%o2 + 0x04]
+	st		%f6, [%o2 + 0x08]
+	st		%f7, [%o2 + 0x0c]
+	retl
+	 VISExit
+ENDPROC(aes_sparc64_decrypt_192)
+
+	.align		32
+ENTRY(aes_sparc64_decrypt_256)
+	/* %o0=key, %o1=input, %o2=output */
+	VISEntry
+	ld		[%o1 + 0x00], %f4
+	ld		[%o1 + 0x04], %f5
+	ld		[%o1 + 0x08], %f6
+	ld		[%o1 + 0x0c], %f7
+	ldd		[%o0 + 0xe0], %f8
+	ldd		[%o0 + 0xe8], %f10
+	ldd		[%o0 + 0xd8], %f12
+	ldd		[%o0 + 0xd0], %f14
+	ldd		[%o0 + 0xc8], %f16
+	fxor		%f8, %f4, %f4
+	ldd		[%o0 + 0xc0], %f18
+	fxor		%f10, %f6, %f6
+	ldd		[%o0 + 0xb8], %f20
+	AES_DROUND23(12, 4, 6, 2)
+	ldd		[%o0 + 0xb0], %f22
+	AES_DROUND01(14, 4, 6, 0)
+	ldd		[%o0 + 0xa8], %f24
+	AES_DROUND23(16, 0, 2, 6)
+	ldd		[%o0 + 0xa0], %f26
+	AES_DROUND01(18, 0, 2, 4)
+	ldd		[%o0 + 0x98], %f12
+	AES_DROUND23(20, 4, 6, 2)
+	ldd		[%o0 + 0x90], %f14
+	AES_DROUND01(22, 4, 6, 0)
+	ldd		[%o0 + 0x88], %f16
+	AES_DROUND23(24, 0, 2, 6)
+	ldd		[%o0 + 0x80], %f18
+	AES_DROUND01(26, 0, 2, 4)
+	ldd		[%o0 + 0x78], %f20
+	AES_DROUND23(12, 4, 6, 2)
+	ldd		[%o0 + 0x70], %f22
+	AES_DROUND01(14, 4, 6, 0)
+	ldd		[%o0 + 0x68], %f24
+	AES_DROUND23(16, 0, 2, 6)
+	ldd		[%o0 + 0x60], %f26
+	AES_DROUND01(18, 0, 2, 4)
+	ldd		[%o0 + 0x58], %f28
+	AES_DROUND23(20, 4, 6, 2)
+	ldd		[%o0 + 0x50], %f30
+	AES_DROUND01(22, 4, 6, 0)
+	ldd		[%o0 + 0x48], %f32
+	AES_DROUND23(24, 0, 2, 6)
+	ldd		[%o0 + 0x40], %f34
+	AES_DROUND01(26, 0, 2, 4)
+	ldd		[%o0 + 0x38], %f36
+	AES_DROUND23(28, 4, 6, 2)
+	ldd		[%o0 + 0x30], %f38
+	AES_DROUND01(30, 4, 6, 0)
+	ldd		[%o0 + 0x28], %f40
+	AES_DROUND23(32, 0, 2, 6)
+	ldd		[%o0 + 0x20], %f42
+	AES_DROUND01(34, 0, 2, 4)
+	ldd		[%o0 + 0x18], %f44
+	AES_DROUND23(36, 4, 6, 2)
+	ldd		[%o0 + 0x10], %f46
+	AES_DROUND01(38, 4, 6, 0)
+	ldd		[%o0 + 0x08], %f48
+	AES_DROUND23(40, 0, 2, 6)
+	ldd		[%o0 + 0x00], %f50
+	AES_DROUND01(42, 0, 2, 4)
+	AES_DROUND23(44, 4, 6, 2)
+	AES_DROUND01(46, 4, 6, 0)
+	AES_DROUND23_L(48, 0, 2, 6)
+	AES_DROUND01_L(50, 0, 2, 4)
+	st		%f4, [%o2 + 0x00]
+	st		%f5, [%o2 + 0x04]
+	st		%f6, [%o2 + 0x08]
+	st		%f7, [%o2 + 0x0c]
+	retl
+	 VISExit
+ENDPROC(aes_sparc64_decrypt_256)
+
+	.align		32
+ENTRY(aes_sparc64_load_encrypt_keys_128)
+	/* %o0=key */
+	VISEntry
+	ldd		[%o0 + 0x10], %f8
+	ldd		[%o0 + 0x18], %f10
+	ldd		[%o0 + 0x20], %f12
+	ldd		[%o0 + 0x28], %f14
+	ldd		[%o0 + 0x30], %f16
+	ldd		[%o0 + 0x38], %f18
+	ldd		[%o0 + 0x40], %f20
+	ldd		[%o0 + 0x48], %f22
+	ldd		[%o0 + 0x50], %f24
+	ldd		[%o0 + 0x58], %f26
+	ldd		[%o0 + 0x60], %f28
+	ldd		[%o0 + 0x68], %f30
+	ldd		[%o0 + 0x70], %f32
+	ldd		[%o0 + 0x78], %f34
+	ldd		[%o0 + 0x80], %f36
+	ldd		[%o0 + 0x88], %f38
+	ldd		[%o0 + 0x90], %f40
+	ldd		[%o0 + 0x98], %f42
+	ldd		[%o0 + 0xa0], %f44
+	retl
+	 ldd		[%o0 + 0xa8], %f46
+ENDPROC(aes_sparc64_load_encrypt_keys_128)
+
+	.align		32
+ENTRY(aes_sparc64_load_encrypt_keys_192)
+	/* %o0=key */
+	VISEntry
+	ldd		[%o0 + 0x10], %f8
+	ldd		[%o0 + 0x18], %f10
+	ldd		[%o0 + 0x20], %f12
+	ldd		[%o0 + 0x28], %f14
+	ldd		[%o0 + 0x30], %f16
+	ldd		[%o0 + 0x38], %f18
+	ldd		[%o0 + 0x40], %f20
+	ldd		[%o0 + 0x48], %f22
+	ldd		[%o0 + 0x50], %f24
+	ldd		[%o0 + 0x58], %f26
+	ldd		[%o0 + 0x60], %f28
+	ldd		[%o0 + 0x68], %f30
+	ldd		[%o0 + 0x70], %f32
+	ldd		[%o0 + 0x78], %f34
+	ldd		[%o0 + 0x80], %f36
+	ldd		[%o0 + 0x88], %f38
+	ldd		[%o0 + 0x90], %f40
+	ldd		[%o0 + 0x98], %f42
+	ldd		[%o0 + 0xa0], %f44
+	ldd		[%o0 + 0xa8], %f46
+	ldd		[%o0 + 0xb0], %f48
+	ldd		[%o0 + 0xb8], %f50
+	ldd		[%o0 + 0xc0], %f52
+	retl
+	 ldd		[%o0 + 0xc8], %f54
+ENDPROC(aes_sparc64_load_encrypt_keys_192)
+
+	.align		32
+ENTRY(aes_sparc64_load_encrypt_keys_256)
+	/* %o0=key */
+	VISEntry
+	ldd		[%o0 + 0x10], %f8
+	ldd		[%o0 + 0x18], %f10
+	ldd		[%o0 + 0x20], %f12
+	ldd		[%o0 + 0x28], %f14
+	ldd		[%o0 + 0x30], %f16
+	ldd		[%o0 + 0x38], %f18
+	ldd		[%o0 + 0x40], %f20
+	ldd		[%o0 + 0x48], %f22
+	ldd		[%o0 + 0x50], %f24
+	ldd		[%o0 + 0x58], %f26
+	ldd		[%o0 + 0x60], %f28
+	ldd		[%o0 + 0x68], %f30
+	ldd		[%o0 + 0x70], %f32
+	ldd		[%o0 + 0x78], %f34
+	ldd		[%o0 + 0x80], %f36
+	ldd		[%o0 + 0x88], %f38
+	ldd		[%o0 + 0x90], %f40
+	ldd		[%o0 + 0x98], %f42
+	ldd		[%o0 + 0xa0], %f44
+	ldd		[%o0 + 0xa8], %f46
+	ldd		[%o0 + 0xb0], %f48
+	ldd		[%o0 + 0xb8], %f50
+	ldd		[%o0 + 0xc0], %f52
+	ldd		[%o0 + 0xc8], %f54
+	ldd		[%o0 + 0xd0], %f56
+	ldd		[%o0 + 0xd8], %f58
+	ldd		[%o0 + 0xe0], %f60
+	retl
+	 ldd		[%o0 + 0xe8], %f62
+ENDPROC(aes_sparc64_load_encrypt_keys_256)
+
+	.align		32
+ENTRY(aes_sparc64_load_decrypt_keys_128)
+	/* %o0=key */
+	VISEntry
+	ldd		[%o0 + 0x98], %f8
+	ldd		[%o0 + 0x90], %f10
+	ldd		[%o0 + 0x88], %f12
+	ldd		[%o0 + 0x80], %f14
+	ldd		[%o0 + 0x78], %f16
+	ldd		[%o0 + 0x70], %f18
+	ldd		[%o0 + 0x68], %f20
+	ldd		[%o0 + 0x60], %f22
+	ldd		[%o0 + 0x58], %f24
+	ldd		[%o0 + 0x50], %f26
+	ldd		[%o0 + 0x48], %f28
+	ldd		[%o0 + 0x40], %f30
+	ldd		[%o0 + 0x38], %f32
+	ldd		[%o0 + 0x30], %f34
+	ldd		[%o0 + 0x28], %f36
+	ldd		[%o0 + 0x20], %f38
+	ldd		[%o0 + 0x18], %f40
+	ldd		[%o0 + 0x10], %f42
+	ldd		[%o0 + 0x08], %f44
+	retl
+	 ldd		[%o0 + 0x00], %f46
+ENDPROC(aes_sparc64_load_decrypt_keys_128)
+
+	.align		32
+ENTRY(aes_sparc64_load_decrypt_keys_192)
+	/* %o0=key */
+	VISEntry
+	ldd		[%o0 + 0xb8], %f8
+	ldd		[%o0 + 0xb0], %f10
+	ldd		[%o0 + 0xa8], %f12
+	ldd		[%o0 + 0xa0], %f14
+	ldd		[%o0 + 0x98], %f16
+	ldd		[%o0 + 0x90], %f18
+	ldd		[%o0 + 0x88], %f20
+	ldd		[%o0 + 0x80], %f22
+	ldd		[%o0 + 0x78], %f24
+	ldd		[%o0 + 0x70], %f26
+	ldd		[%o0 + 0x68], %f28
+	ldd		[%o0 + 0x60], %f30
+	ldd		[%o0 + 0x58], %f32
+	ldd		[%o0 + 0x50], %f34
+	ldd		[%o0 + 0x48], %f36
+	ldd		[%o0 + 0x40], %f38
+	ldd		[%o0 + 0x38], %f40
+	ldd		[%o0 + 0x30], %f42
+	ldd		[%o0 + 0x28], %f44
+	ldd		[%o0 + 0x20], %f46
+	ldd		[%o0 + 0x18], %f48
+	ldd		[%o0 + 0x10], %f50
+	ldd		[%o0 + 0x08], %f52
+	retl
+	 ldd		[%o0 + 0x00], %f54
+ENDPROC(aes_sparc64_load_decrypt_keys_192)
+
+	.align		32
+ENTRY(aes_sparc64_load_decrypt_keys_256)
+	/* %o0=key */
+	VISEntry
+	ldd		[%o0 + 0xd8], %f8
+	ldd		[%o0 + 0xd0], %f10
+	ldd		[%o0 + 0xc8], %f12
+	ldd		[%o0 + 0xc0], %f14
+	ldd		[%o0 + 0xb8], %f16
+	ldd		[%o0 + 0xb0], %f18
+	ldd		[%o0 + 0xa8], %f20
+	ldd		[%o0 + 0xa0], %f22
+	ldd		[%o0 + 0x98], %f24
+	ldd		[%o0 + 0x90], %f26
+	ldd		[%o0 + 0x88], %f28
+	ldd		[%o0 + 0x80], %f30
+	ldd		[%o0 + 0x78], %f32
+	ldd		[%o0 + 0x70], %f34
+	ldd		[%o0 + 0x68], %f36
+	ldd		[%o0 + 0x60], %f38
+	ldd		[%o0 + 0x58], %f40
+	ldd		[%o0 + 0x50], %f42
+	ldd		[%o0 + 0x48], %f44
+	ldd		[%o0 + 0x40], %f46
+	ldd		[%o0 + 0x38], %f48
+	ldd		[%o0 + 0x30], %f50
+	ldd		[%o0 + 0x28], %f52
+	ldd		[%o0 + 0x20], %f54
+	ldd		[%o0 + 0x18], %f56
+	ldd		[%o0 + 0x10], %f58
+	ldd		[%o0 + 0x08], %f60
+	retl
+	 ldd		[%o0 + 0x00], %f62
+ENDPROC(aes_sparc64_load_decrypt_keys_256)
+
+	.align		32
+ENTRY(aes_sparc64_ecb_encrypt_128)
+	/* %o0=key, %o1=input, %o2=output, %o3=len */
+	ldx		[%o0 + 0x00], %g1
+	subcc		%o3, 0x10, %o3
+	be		10f
+	 ldx		[%o0 + 0x08], %g2
+1:	ldx		[%o1 + 0x00], %g3
+	ldx		[%o1 + 0x08], %g7
+	ldx		[%o1 + 0x10], %o4
+	ldx		[%o1 + 0x18], %o5
+	xor		%g1, %g3, %g3
+	xor		%g2, %g7, %g7
+	MOVXTOD_G3_F4
+	MOVXTOD_G7_F6
+	xor		%g1, %o4, %g3
+	xor		%g2, %o5, %g7
+	MOVXTOD_G3_F60
+	MOVXTOD_G7_F62
+	ENCRYPT_128_2(8, 4, 6, 60, 62, 0, 2, 56, 58)
+	std		%f4, [%o2 + 0x00]
+	std		%f6, [%o2 + 0x08]
+	std		%f60, [%o2 + 0x10]
+	std		%f62, [%o2 + 0x18]
+	sub		%o3, 0x20, %o3
+	add		%o1, 0x20, %o1
+	brgz		%o3, 1b
+	 add		%o2, 0x20, %o2
+	brlz,pt		%o3, 11f
+	 nop
+10:	ldx		[%o1 + 0x00], %g3
+	ldx		[%o1 + 0x08], %g7
+	xor		%g1, %g3, %g3
+	xor		%g2, %g7, %g7
+	MOVXTOD_G3_F4
+	MOVXTOD_G7_F6
+	ENCRYPT_128(8, 4, 6, 0, 2)
+	std		%f4, [%o2 + 0x00]
+	std		%f6, [%o2 + 0x08]
+11:	retl
+	 nop
+ENDPROC(aes_sparc64_ecb_encrypt_128)
+
+	.align		32
+ENTRY(aes_sparc64_ecb_encrypt_192)
+	/* %o0=key, %o1=input, %o2=output, %o3=len */
+	ldx		[%o0 + 0x00], %g1
+	subcc		%o3, 0x10, %o3
+	be		10f
+	 ldx		[%o0 + 0x08], %g2
+1:	ldx		[%o1 + 0x00], %g3
+	ldx		[%o1 + 0x08], %g7
+	ldx		[%o1 + 0x10], %o4
+	ldx		[%o1 + 0x18], %o5
+	xor		%g1, %g3, %g3
+	xor		%g2, %g7, %g7
+	MOVXTOD_G3_F4
+	MOVXTOD_G7_F6
+	xor		%g1, %o4, %g3
+	xor		%g2, %o5, %g7
+	MOVXTOD_G3_F60
+	MOVXTOD_G7_F62
+	ENCRYPT_192_2(8, 4, 6, 60, 62, 0, 2, 56, 58)
+	std		%f4, [%o2 + 0x00]
+	std		%f6, [%o2 + 0x08]
+	std		%f60, [%o2 + 0x10]
+	std		%f62, [%o2 + 0x18]
+	sub		%o3, 0x20, %o3
+	add		%o1, 0x20, %o1
+	brgz		%o3, 1b
+	 add		%o2, 0x20, %o2
+	brlz,pt		%o3, 11f
+	 nop
+10:	ldx		[%o1 + 0x00], %g3
+	ldx		[%o1 + 0x08], %g7
+	xor		%g1, %g3, %g3
+	xor		%g2, %g7, %g7
+	MOVXTOD_G3_F4
+	MOVXTOD_G7_F6
+	ENCRYPT_192(8, 4, 6, 0, 2)
+	std		%f4, [%o2 + 0x00]
+	std		%f6, [%o2 + 0x08]
+11:	retl
+	 nop
+ENDPROC(aes_sparc64_ecb_encrypt_192)
+
+	.align		32
+ENTRY(aes_sparc64_ecb_encrypt_256)
+	/* %o0=key, %o1=input, %o2=output, %o3=len */
+	ldx		[%o0 + 0x00], %g1
+	subcc		%o3, 0x10, %o3
+	be		10f
+	 ldx		[%o0 + 0x08], %g2
+1:	ldx		[%o1 + 0x00], %g3
+	ldx		[%o1 + 0x08], %g7
+	ldx		[%o1 + 0x10], %o4
+	ldx		[%o1 + 0x18], %o5
+	xor		%g1, %g3, %g3
+	xor		%g2, %g7, %g7
+	MOVXTOD_G3_F4
+	MOVXTOD_G7_F6
+	xor		%g1, %o4, %g3
+	xor		%g2, %o5, %g7
+	MOVXTOD_G3_F0
+	MOVXTOD_G7_F2
+	ENCRYPT_256_2(8, 4, 6, 0, 2)
+	std		%f4, [%o2 + 0x00]
+	std		%f6, [%o2 + 0x08]
+	std		%f0, [%o2 + 0x10]
+	std		%f2, [%o2 + 0x18]
+	sub		%o3, 0x20, %o3
+	add		%o1, 0x20, %o1
+	brgz		%o3, 1b
+	 add		%o2, 0x20, %o2
+	brlz,pt		%o3, 11f
+	 nop
+10:	ldd		[%o0 + 0xd0], %f56
+	ldd		[%o0 + 0xd8], %f58
+	ldd		[%o0 + 0xe0], %f60
+	ldd		[%o0 + 0xe8], %f62
+	ldx		[%o1 + 0x00], %g3
+	ldx		[%o1 + 0x08], %g7
+	xor		%g1, %g3, %g3
+	xor		%g2, %g7, %g7
+	MOVXTOD_G3_F4
+	MOVXTOD_G7_F6
+	ENCRYPT_256(8, 4, 6, 0, 2)
+	std		%f4, [%o2 + 0x00]
+	std		%f6, [%o2 + 0x08]
+11:	retl
+	 nop
+ENDPROC(aes_sparc64_ecb_encrypt_256)
+
+	.align		32
+ENTRY(aes_sparc64_ecb_decrypt_128)
+	/* %o0=&key[key_len], %o1=input, %o2=output, %o3=len */
+	ldx		[%o0 - 0x10], %g1
+	subcc		%o3, 0x10, %o3
+	be		10f
+	 ldx		[%o0 - 0x08], %g2
+1:	ldx		[%o1 + 0x00], %g3
+	ldx		[%o1 + 0x08], %g7
+	ldx		[%o1 + 0x10], %o4
+	ldx		[%o1 + 0x18], %o5
+	xor		%g1, %g3, %g3
+	xor		%g2, %g7, %g7
+	MOVXTOD_G3_F4
+	MOVXTOD_G7_F6
+	xor		%g1, %o4, %g3
+	xor		%g2, %o5, %g7
+	MOVXTOD_G3_F60
+	MOVXTOD_G7_F62
+	DECRYPT_128_2(8, 4, 6, 60, 62, 0, 2, 56, 58)
+	std		%f4, [%o2 + 0x00]
+	std		%f6, [%o2 + 0x08]
+	std		%f60, [%o2 + 0x10]
+	std		%f62, [%o2 + 0x18]
+	sub		%o3, 0x20, %o3
+	add		%o1, 0x20, %o1
+	brgz,pt		%o3, 1b
+	 add		%o2, 0x20, %o2
+	brlz,pt		%o3, 11f
+	 nop
+10:	ldx		[%o1 + 0x00], %g3
+	ldx		[%o1 + 0x08], %g7
+	xor		%g1, %g3, %g3
+	xor		%g2, %g7, %g7
+	MOVXTOD_G3_F4
+	MOVXTOD_G7_F6
+	DECRYPT_128(8, 4, 6, 0, 2)
+	std		%f4, [%o2 + 0x00]
+	std		%f6, [%o2 + 0x08]
+11:	retl
+	 nop
+ENDPROC(aes_sparc64_ecb_decrypt_128)
+
+	.align		32
+ENTRY(aes_sparc64_ecb_decrypt_192)
+	/* %o0=&key[key_len], %o1=input, %o2=output, %o3=len */
+	ldx		[%o0 - 0x10], %g1
+	subcc		%o3, 0x10, %o3
+	be		10f
+	 ldx		[%o0 - 0x08], %g2
+1:	ldx		[%o1 + 0x00], %g3
+	ldx		[%o1 + 0x08], %g7
+	ldx		[%o1 + 0x10], %o4
+	ldx		[%o1 + 0x18], %o5
+	xor		%g1, %g3, %g3
+	xor		%g2, %g7, %g7
+	MOVXTOD_G3_F4
+	MOVXTOD_G7_F6
+	xor		%g1, %o4, %g3
+	xor		%g2, %o5, %g7
+	MOVXTOD_G3_F60
+	MOVXTOD_G7_F62
+	DECRYPT_192_2(8, 4, 6, 60, 62, 0, 2, 56, 58)
+	std		%f4, [%o2 + 0x00]
+	std		%f6, [%o2 + 0x08]
+	std		%f60, [%o2 + 0x10]
+	std		%f62, [%o2 + 0x18]
+	sub		%o3, 0x20, %o3
+	add		%o1, 0x20, %o1
+	brgz,pt		%o3, 1b
+	 add		%o2, 0x20, %o2
+	brlz,pt		%o3, 11f
+	 nop
+10:	ldx		[%o1 + 0x00], %g3
+	ldx		[%o1 + 0x08], %g7
+	xor		%g1, %g3, %g3
+	xor		%g2, %g7, %g7
+	MOVXTOD_G3_F4
+	MOVXTOD_G7_F6
+	DECRYPT_192(8, 4, 6, 0, 2)
+	std		%f4, [%o2 + 0x00]
+	std		%f6, [%o2 + 0x08]
+11:	retl
+	 nop
+ENDPROC(aes_sparc64_ecb_decrypt_192)
+
+	.align		32
+ENTRY(aes_sparc64_ecb_decrypt_256)
+	/* %o0=&key[key_len], %o1=input, %o2=output, %o3=len */
+	ldx		[%o0 - 0x10], %g1
+	subcc		%o3, 0x10, %o3
+	ldx		[%o0 - 0x08], %g2
+	be		10f
+	 sub		%o0, 0xf0, %o0
+1:	ldx		[%o1 + 0x00], %g3
+	ldx		[%o1 + 0x08], %g7
+	ldx		[%o1 + 0x10], %o4
+	ldx		[%o1 + 0x18], %o5
+	xor		%g1, %g3, %g3
+	xor		%g2, %g7, %g7
+	MOVXTOD_G3_F4
+	MOVXTOD_G7_F6
+	xor		%g1, %o4, %g3
+	xor		%g2, %o5, %g7
+	MOVXTOD_G3_F0
+	MOVXTOD_G7_F2
+	DECRYPT_256_2(8, 4, 6, 0, 2)
+	std		%f4, [%o2 + 0x00]
+	std		%f6, [%o2 + 0x08]
+	std		%f0, [%o2 + 0x10]
+	std		%f2, [%o2 + 0x18]
+	sub		%o3, 0x20, %o3
+	add		%o1, 0x20, %o1
+	brgz,pt		%o3, 1b
+	 add		%o2, 0x20, %o2
+	brlz,pt		%o3, 11f
+	 nop
+10:	ldd		[%o0 + 0x18], %f56
+	ldd		[%o0 + 0x10], %f58
+	ldd		[%o0 + 0x08], %f60
+	ldd		[%o0 + 0x00], %f62
+	ldx		[%o1 + 0x00], %g3
+	ldx		[%o1 + 0x08], %g7
+	xor		%g1, %g3, %g3
+	xor		%g2, %g7, %g7
+	MOVXTOD_G3_F4
+	MOVXTOD_G7_F6
+	DECRYPT_256(8, 4, 6, 0, 2)
+	std		%f4, [%o2 + 0x00]
+	std		%f6, [%o2 + 0x08]
+11:	retl
+	 nop
+ENDPROC(aes_sparc64_ecb_decrypt_256)
+
+	.align		32
+ENTRY(aes_sparc64_cbc_encrypt_128)
+	/* %o0=key, %o1=input, %o2=output, %o3=len, %o4=IV */
+	ldd		[%o4 + 0x00], %f4
+	ldd		[%o4 + 0x08], %f6
+	ldx		[%o0 + 0x00], %g1
+	ldx		[%o0 + 0x08], %g2
+1:	ldx		[%o1 + 0x00], %g3
+	ldx		[%o1 + 0x08], %g7
+	add		%o1, 0x10, %o1
+	xor		%g1, %g3, %g3
+	xor		%g2, %g7, %g7
+	MOVXTOD_G3_F0
+	MOVXTOD_G7_F2
+	fxor		%f4, %f0, %f4
+	fxor		%f6, %f2, %f6
+	ENCRYPT_128(8, 4, 6, 0, 2)
+	std		%f4, [%o2 + 0x00]
+	std		%f6, [%o2 + 0x08]
+	subcc		%o3, 0x10, %o3
+	bne,pt		%xcc, 1b
+	 add		%o2, 0x10, %o2
+	std		%f4, [%o4 + 0x00]
+	std		%f6, [%o4 + 0x08]
+	retl
+	 nop
+ENDPROC(aes_sparc64_cbc_encrypt_128)
+
+	.align		32
+ENTRY(aes_sparc64_cbc_encrypt_192)
+	/* %o0=key, %o1=input, %o2=output, %o3=len, %o4=IV */
+	ldd		[%o4 + 0x00], %f4
+	ldd		[%o4 + 0x08], %f6
+	ldx		[%o0 + 0x00], %g1
+	ldx		[%o0 + 0x08], %g2
+1:	ldx		[%o1 + 0x00], %g3
+	ldx		[%o1 + 0x08], %g7
+	add		%o1, 0x10, %o1
+	xor		%g1, %g3, %g3
+	xor		%g2, %g7, %g7
+	MOVXTOD_G3_F0
+	MOVXTOD_G7_F2
+	fxor		%f4, %f0, %f4
+	fxor		%f6, %f2, %f6
+	ENCRYPT_192(8, 4, 6, 0, 2)
+	std		%f4, [%o2 + 0x00]
+	std		%f6, [%o2 + 0x08]
+	subcc		%o3, 0x10, %o3
+	bne,pt		%xcc, 1b
+	 add		%o2, 0x10, %o2
+	std		%f4, [%o4 + 0x00]
+	std		%f6, [%o4 + 0x08]
+	retl
+	 nop
+ENDPROC(aes_sparc64_cbc_encrypt_192)
+
+	.align		32
+ENTRY(aes_sparc64_cbc_encrypt_256)
+	/* %o0=key, %o1=input, %o2=output, %o3=len, %o4=IV */
+	ldd		[%o4 + 0x00], %f4
+	ldd		[%o4 + 0x08], %f6
+	ldx		[%o0 + 0x00], %g1
+	ldx		[%o0 + 0x08], %g2
+1:	ldx		[%o1 + 0x00], %g3
+	ldx		[%o1 + 0x08], %g7
+	add		%o1, 0x10, %o1
+	xor		%g1, %g3, %g3
+	xor		%g2, %g7, %g7
+	MOVXTOD_G3_F0
+	MOVXTOD_G7_F2
+	fxor		%f4, %f0, %f4
+	fxor		%f6, %f2, %f6
+	ENCRYPT_256(8, 4, 6, 0, 2)
+	std		%f4, [%o2 + 0x00]
+	std		%f6, [%o2 + 0x08]
+	subcc		%o3, 0x10, %o3
+	bne,pt		%xcc, 1b
+	 add		%o2, 0x10, %o2
+	std		%f4, [%o4 + 0x00]
+	std		%f6, [%o4 + 0x08]
+	retl
+	 nop
+ENDPROC(aes_sparc64_cbc_encrypt_256)
+
+	.align		32
+ENTRY(aes_sparc64_cbc_decrypt_128)
+	/* %o0=&key[key_len], %o1=input, %o2=output, %o3=len, %o4=iv */
+	ldx		[%o0 - 0x10], %g1
+	ldx		[%o0 - 0x08], %g2
+	ldx		[%o4 + 0x00], %o0
+	ldx		[%o4 + 0x08], %o5
+1:	ldx		[%o1 + 0x00], %g3
+	ldx		[%o1 + 0x08], %g7
+	add		%o1, 0x10, %o1
+	xor		%g1, %g3, %g3
+	xor		%g2, %g7, %g7
+	MOVXTOD_G3_F4
+	MOVXTOD_G7_F6
+	DECRYPT_128(8, 4, 6, 0, 2)
+	MOVXTOD_O0_F0
+	MOVXTOD_O5_F2
+	xor		%g1, %g3, %o0
+	xor		%g2, %g7, %o5
+	fxor		%f4, %f0, %f4
+	fxor		%f6, %f2, %f6
+	std		%f4, [%o2 + 0x00]
+	std		%f6, [%o2 + 0x08]
+	subcc		%o3, 0x10, %o3
+	bne,pt		%xcc, 1b
+	 add		%o2, 0x10, %o2
+	stx		%o0, [%o4 + 0x00]
+	stx		%o5, [%o4 + 0x08]
+	retl
+	 nop
+ENDPROC(aes_sparc64_cbc_decrypt_128)
+
+	.align		32
+ENTRY(aes_sparc64_cbc_decrypt_192)
+	/* %o0=&key[key_len], %o1=input, %o2=output, %o3=len, %o4=iv */
+	ldx		[%o0 - 0x10], %g1
+	ldx		[%o0 - 0x08], %g2
+	ldx		[%o4 + 0x00], %o0
+	ldx		[%o4 + 0x08], %o5
+1:	ldx		[%o1 + 0x00], %g3
+	ldx		[%o1 + 0x08], %g7
+	add		%o1, 0x10, %o1
+	xor		%g1, %g3, %g3
+	xor		%g2, %g7, %g7
+	MOVXTOD_G3_F4
+	MOVXTOD_G7_F6
+	DECRYPT_192(8, 4, 6, 0, 2)
+	MOVXTOD_O0_F0
+	MOVXTOD_O5_F2
+	xor		%g1, %g3, %o0
+	xor		%g2, %g7, %o5
+	fxor		%f4, %f0, %f4
+	fxor		%f6, %f2, %f6
+	std		%f4, [%o2 + 0x00]
+	std		%f6, [%o2 + 0x08]
+	subcc		%o3, 0x10, %o3
+	bne,pt		%xcc, 1b
+	 add		%o2, 0x10, %o2
+	stx		%o0, [%o4 + 0x00]
+	stx		%o5, [%o4 + 0x08]
+	retl
+	 nop
+ENDPROC(aes_sparc64_cbc_decrypt_192)
+
+	.align		32
+ENTRY(aes_sparc64_cbc_decrypt_256)
+	/* %o0=&key[key_len], %o1=input, %o2=output, %o3=len, %o4=iv */
+	ldx		[%o0 - 0x10], %g1
+	ldx		[%o0 - 0x08], %g2
+	ldx		[%o4 + 0x00], %o0
+	ldx		[%o4 + 0x08], %o5
+1:	ldx		[%o1 + 0x00], %g3
+	ldx		[%o1 + 0x08], %g7
+	add		%o1, 0x10, %o1
+	xor		%g1, %g3, %g3
+	xor		%g2, %g7, %g7
+	MOVXTOD_G3_F4
+	MOVXTOD_G7_F6
+	DECRYPT_256(8, 4, 6, 0, 2)
+	MOVXTOD_O0_F0
+	MOVXTOD_O5_F2
+	xor		%g1, %g3, %o0
+	xor		%g2, %g7, %o5
+	fxor		%f4, %f0, %f4
+	fxor		%f6, %f2, %f6
+	std		%f4, [%o2 + 0x00]
+	std		%f6, [%o2 + 0x08]
+	subcc		%o3, 0x10, %o3
+	bne,pt		%xcc, 1b
+	 add		%o2, 0x10, %o2
+	stx		%o0, [%o4 + 0x00]
+	stx		%o5, [%o4 + 0x08]
+	retl
+	 nop
+ENDPROC(aes_sparc64_cbc_decrypt_256)
+
+	.align		32
+ENTRY(aes_sparc64_ctr_crypt_128)
+	/* %o0=key, %o1=input, %o2=output, %o3=len, %o4=IV */
+	ldx		[%o4 + 0x00], %g3
+	ldx		[%o4 + 0x08], %g7
+	subcc		%o3, 0x10, %o3
+	ldx		[%o0 + 0x00], %g1
+	be		10f
+	 ldx		[%o0 + 0x08], %g2
+1:	xor		%g1, %g3, %o5
+	MOVXTOD_O5_F0
+	xor		%g2, %g7, %o5
+	MOVXTOD_O5_F2
+	add		%g7, 1, %g7
+	add		%g3, 1, %o5
+	movrz		%g7, %o5, %g3
+	xor		%g1, %g3, %o5
+	MOVXTOD_O5_F4
+	xor		%g2, %g7, %o5
+	MOVXTOD_O5_F6
+	add		%g7, 1, %g7
+	add		%g3, 1, %o5
+	movrz		%g7, %o5, %g3
+	ENCRYPT_128_2(8, 0, 2, 4, 6, 56, 58, 60, 62)
+	ldd		[%o1 + 0x00], %f56
+	ldd		[%o1 + 0x08], %f58
+	ldd		[%o1 + 0x10], %f60
+	ldd		[%o1 + 0x18], %f62
+	fxor		%f56, %f0, %f56
+	fxor		%f58, %f2, %f58
+	fxor		%f60, %f4, %f60
+	fxor		%f62, %f6, %f62
+	std		%f56, [%o2 + 0x00]
+	std		%f58, [%o2 + 0x08]
+	std		%f60, [%o2 + 0x10]
+	std		%f62, [%o2 + 0x18]
+	subcc		%o3, 0x20, %o3
+	add		%o1, 0x20, %o1
+	brgz		%o3, 1b
+	 add		%o2, 0x20, %o2
+	brlz,pt		%o3, 11f
+	 nop
+10:	xor		%g1, %g3, %o5
+	MOVXTOD_O5_F0
+	xor		%g2, %g7, %o5
+	MOVXTOD_O5_F2
+	add		%g7, 1, %g7
+	add		%g3, 1, %o5
+	movrz		%g7, %o5, %g3
+	ENCRYPT_128(8, 0, 2, 4, 6)
+	ldd		[%o1 + 0x00], %f4
+	ldd		[%o1 + 0x08], %f6
+	fxor		%f4, %f0, %f4
+	fxor		%f6, %f2, %f6
+	std		%f4, [%o2 + 0x00]
+	std		%f6, [%o2 + 0x08]
+11:	stx		%g3, [%o4 + 0x00]
+	retl
+	 stx		%g7, [%o4 + 0x08]
+ENDPROC(aes_sparc64_ctr_crypt_128)
+
+	.align		32
+ENTRY(aes_sparc64_ctr_crypt_192)
+	/* %o0=key, %o1=input, %o2=output, %o3=len, %o4=IV */
+	ldx		[%o4 + 0x00], %g3
+	ldx		[%o4 + 0x08], %g7
+	subcc		%o3, 0x10, %o3
+	ldx		[%o0 + 0x00], %g1
+	be		10f
+	 ldx		[%o0 + 0x08], %g2
+1:	xor		%g1, %g3, %o5
+	MOVXTOD_O5_F0
+	xor		%g2, %g7, %o5
+	MOVXTOD_O5_F2
+	add		%g7, 1, %g7
+	add		%g3, 1, %o5
+	movrz		%g7, %o5, %g3
+	xor		%g1, %g3, %o5
+	MOVXTOD_O5_F4
+	xor		%g2, %g7, %o5
+	MOVXTOD_O5_F6
+	add		%g7, 1, %g7
+	add		%g3, 1, %o5
+	movrz		%g7, %o5, %g3
+	ENCRYPT_192_2(8, 0, 2, 4, 6, 56, 58, 60, 62)
+	ldd		[%o1 + 0x00], %f56
+	ldd		[%o1 + 0x08], %f58
+	ldd		[%o1 + 0x10], %f60
+	ldd		[%o1 + 0x18], %f62
+	fxor		%f56, %f0, %f56
+	fxor		%f58, %f2, %f58
+	fxor		%f60, %f4, %f60
+	fxor		%f62, %f6, %f62
+	std		%f56, [%o2 + 0x00]
+	std		%f58, [%o2 + 0x08]
+	std		%f60, [%o2 + 0x10]
+	std		%f62, [%o2 + 0x18]
+	subcc		%o3, 0x20, %o3
+	add		%o1, 0x20, %o1
+	brgz		%o3, 1b
+	 add		%o2, 0x20, %o2
+	brlz,pt		%o3, 11f
+	 nop
+10:	xor		%g1, %g3, %o5
+	MOVXTOD_O5_F0
+	xor		%g2, %g7, %o5
+	MOVXTOD_O5_F2
+	add		%g7, 1, %g7
+	add		%g3, 1, %o5
+	movrz		%g7, %o5, %g3
+	ENCRYPT_192(8, 0, 2, 4, 6)
+	ldd		[%o1 + 0x00], %f4
+	ldd		[%o1 + 0x08], %f6
+	fxor		%f4, %f0, %f4
+	fxor		%f6, %f2, %f6
+	std		%f4, [%o2 + 0x00]
+	std		%f6, [%o2 + 0x08]
+11:	stx		%g3, [%o4 + 0x00]
+	retl
+	 stx		%g7, [%o4 + 0x08]
+ENDPROC(aes_sparc64_ctr_crypt_192)
+
+	.align		32
+ENTRY(aes_sparc64_ctr_crypt_256)
+	/* %o0=key, %o1=input, %o2=output, %o3=len, %o4=IV */
+	ldx		[%o4 + 0x00], %g3
+	ldx		[%o4 + 0x08], %g7
+	subcc		%o3, 0x10, %o3
+	ldx		[%o0 + 0x00], %g1
+	be		10f
+	 ldx		[%o0 + 0x08], %g2
+1:	xor		%g1, %g3, %o5
+	MOVXTOD_O5_F0
+	xor		%g2, %g7, %o5
+	MOVXTOD_O5_F2
+	add		%g7, 1, %g7
+	add		%g3, 1, %o5
+	movrz		%g7, %o5, %g3
+	xor		%g1, %g3, %o5
+	MOVXTOD_O5_F4
+	xor		%g2, %g7, %o5
+	MOVXTOD_O5_F6
+	add		%g7, 1, %g7
+	add		%g3, 1, %o5
+	movrz		%g7, %o5, %g3
+	ENCRYPT_256_2(8, 0, 2, 4, 6)
+	ldd		[%o1 + 0x00], %f56
+	ldd		[%o1 + 0x08], %f58
+	ldd		[%o1 + 0x10], %f60
+	ldd		[%o1 + 0x18], %f62
+	fxor		%f56, %f0, %f56
+	fxor		%f58, %f2, %f58
+	fxor		%f60, %f4, %f60
+	fxor		%f62, %f6, %f62
+	std		%f56, [%o2 + 0x00]
+	std		%f58, [%o2 + 0x08]
+	std		%f60, [%o2 + 0x10]
+	std		%f62, [%o2 + 0x18]
+	subcc		%o3, 0x20, %o3
+	add		%o1, 0x20, %o1
+	brgz		%o3, 1b
+	 add		%o2, 0x20, %o2
+	brlz,pt		%o3, 11f
+	 nop
+10:	ldd		[%o0 + 0xd0], %f56
+	ldd		[%o0 + 0xd8], %f58
+	ldd		[%o0 + 0xe0], %f60
+	ldd		[%o0 + 0xe8], %f62
+	xor		%g1, %g3, %o5
+	MOVXTOD_O5_F0
+	xor		%g2, %g7, %o5
+	MOVXTOD_O5_F2
+	add		%g7, 1, %g7
+	add		%g3, 1, %o5
+	movrz		%g7, %o5, %g3
+	ENCRYPT_256(8, 0, 2, 4, 6)
+	ldd		[%o1 + 0x00], %f4
+	ldd		[%o1 + 0x08], %f6
+	fxor		%f4, %f0, %f4
+	fxor		%f6, %f2, %f6
+	std		%f4, [%o2 + 0x00]
+	std		%f6, [%o2 + 0x08]
+11:	stx		%g3, [%o4 + 0x00]
+	retl
+	 stx		%g7, [%o4 + 0x08]
+ENDPROC(aes_sparc64_ctr_crypt_256)