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, 398 insertions, 0 deletions

diff --git a/arch/alpha/lib/ev6-stxncpy.S b/arch/alpha/lib/ev6-stxncpy.S
new file mode 100644
index 000000000..76da20528
--- /dev/null
+++ b/arch/alpha/lib/ev6-stxncpy.S
@@ -0,0 +1,398 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * arch/alpha/lib/ev6-stxncpy.S
+ * 21264 version contributed by Rick Gorton <rick.gorton@api-networks.com>
+ *
+ * Copy no more than COUNT bytes of the null-terminated string from
+ * SRC to DST.
+ *
+ * This is an internal routine used by strncpy, stpncpy, and strncat.
+ * As such, it uses special linkage conventions to make implementation
+ * of these public functions more efficient.
+ *
+ * On input:
+ *	t9 = return address
+ *	a0 = DST
+ *	a1 = SRC
+ *	a2 = COUNT
+ *
+ * Furthermore, COUNT may not be zero.
+ *
+ * On output:
+ *	t0  = last word written
+ *	t10 = bitmask (with one bit set) indicating the byte position of
+ *	      the end of the range specified by COUNT
+ *	t12 = bitmask (with one bit set) indicating the last byte written
+ *	a0  = unaligned address of the last *word* written
+ *	a2  = the number of full words left in COUNT
+ *
+ * Furthermore, v0, a3-a5, t11, and $at are untouched.
+ *
+ * Much of the information about 21264 scheduling/coding comes from:
+ *	Compiler Writer's Guide for the Alpha 21264
+ *	abbreviated as 'CWG' in other comments here
+ *	ftp.digital.com/pub/Digital/info/semiconductor/literature/dsc-library.html
+ * Scheduling notation:
+ *	E	- either cluster
+ *	U	- upper subcluster; U0 - subcluster U0; U1 - subcluster U1
+ *	L	- lower subcluster; L0 - subcluster L0; L1 - subcluster L1
+ * Try not to change the actual algorithm if possible for consistency.
+ */
+
+#include <asm/regdef.h>
+
+	.set noat
+	.set noreorder
+
+	.text
+
+/* There is a problem with either gdb (as of 4.16) or gas (as of 2.7) that
+   doesn't like putting the entry point for a procedure somewhere in the
+   middle of the procedure descriptor.  Work around this by putting the
+   aligned copy in its own procedure descriptor */
+
+
+	.ent stxncpy_aligned
+	.align 4
+stxncpy_aligned:
+	.frame sp, 0, t9, 0
+	.prologue 0
+
+	/* On entry to this basic block:
+	   t0 == the first destination word for masking back in
+	   t1 == the first source word.  */
+
+	/* Create the 1st output word and detect 0's in the 1st input word.  */
+	lda	t2, -1		# E : build a mask against false zero
+	mskqh	t2, a1, t2	# U :   detection in the src word (stall)
+	mskqh	t1, a1, t3	# U :
+	ornot	t1, t2, t2	# E : (stall)
+
+	mskql	t0, a1, t0	# U : assemble the first output word
+	cmpbge	zero, t2, t8	# E : bits set iff null found
+	or	t0, t3, t0	# E : (stall)
+	beq	a2, $a_eoc	# U :
+
+	bne	t8, $a_eos	# U :
+	nop
+	nop
+	nop
+
+	/* On entry to this basic block:
+	   t0 == a source word not containing a null.  */
+
+	/*
+	 * nops here to:
+	 *	separate store quads from load quads
+	 *	limit of 1 bcond/quad to permit training
+	 */
+$a_loop:
+	stq_u	t0, 0(a0)	# L :
+	addq	a0, 8, a0	# E :
+	subq	a2, 1, a2	# E :
+	nop
+
+	ldq_u	t0, 0(a1)	# L :
+	addq	a1, 8, a1	# E :
+	cmpbge	zero, t0, t8	# E :
+	beq	a2, $a_eoc      # U :
+
+	beq	t8, $a_loop	# U :
+	nop
+	nop
+	nop
+
+	/* Take care of the final (partial) word store.  At this point
+	   the end-of-count bit is set in t8 iff it applies.
+
+	   On entry to this basic block we have:
+	   t0 == the source word containing the null
+	   t8 == the cmpbge mask that found it.  */
+
+$a_eos:
+	negq	t8, t12		# E : find low bit set
+	and	t8, t12, t12	# E : (stall)
+	/* For the sake of the cache, don't read a destination word
+	   if we're not going to need it.  */
+	and	t12, 0x80, t6	# E : (stall)
+	bne	t6, 1f		# U : (stall)
+
+	/* We're doing a partial word store and so need to combine
+	   our source and original destination words.  */
+	ldq_u	t1, 0(a0)	# L :
+	subq	t12, 1, t6	# E :
+	or	t12, t6, t8	# E : (stall)
+	zapnot	t0, t8, t0	# U : clear src bytes > null (stall)
+
+	zap	t1, t8, t1	# .. e1 : clear dst bytes <= null
+	or	t0, t1, t0	# e1    : (stall)
+	nop
+	nop
+
+1:	stq_u	t0, 0(a0)	# L :
+	ret	(t9)		# L0 : Latency=3
+	nop
+	nop
+
+	/* Add the end-of-count bit to the eos detection bitmask.  */
+$a_eoc:
+	or	t10, t8, t8	# E :
+	br	$a_eos		# L0 : Latency=3
+	nop
+	nop
+
+	.end stxncpy_aligned
+
+	.align 4
+	.ent __stxncpy
+	.globl __stxncpy
+__stxncpy:
+	.frame sp, 0, t9, 0
+	.prologue 0
+
+	/* Are source and destination co-aligned?  */
+	xor	a0, a1, t1	# E :
+	and	a0, 7, t0	# E : find dest misalignment
+	and	t1, 7, t1	# E : (stall)
+	addq	a2, t0, a2	# E : bias count by dest misalignment (stall)
+
+	subq	a2, 1, a2	# E :
+	and	a2, 7, t2	# E : (stall)
+	srl	a2, 3, a2	# U : a2 = loop counter = (count - 1)/8 (stall)
+	addq	zero, 1, t10	# E :
+
+	sll	t10, t2, t10	# U : t10 = bitmask of last count byte
+	bne	t1, $unaligned	# U :
+	/* We are co-aligned; take care of a partial first word.  */
+	ldq_u	t1, 0(a1)	# L : load first src word
+	addq	a1, 8, a1	# E :
+
+	beq	t0, stxncpy_aligned     # U : avoid loading dest word if not needed
+	ldq_u	t0, 0(a0)	# L :
+	nop
+	nop
+
+	br	stxncpy_aligned	# .. e1 :
+	nop
+	nop
+	nop
+
+
+
+/* The source and destination are not co-aligned.  Align the destination
+   and cope.  We have to be very careful about not reading too much and
+   causing a SEGV.  */
+
+	.align 4
+$u_head:
+	/* We know just enough now to be able to assemble the first
+	   full source word.  We can still find a zero at the end of it
+	   that prevents us from outputting the whole thing.
+
+	   On entry to this basic block:
+	   t0 == the first dest word, unmasked
+	   t1 == the shifted low bits of the first source word
+	   t6 == bytemask that is -1 in dest word bytes */
+
+	ldq_u	t2, 8(a1)	# L : Latency=3 load second src word
+	addq	a1, 8, a1	# E :
+	mskql	t0, a0, t0	# U : mask trailing garbage in dst
+	extqh	t2, a1, t4	# U : (3 cycle stall on t2)
+
+	or	t1, t4, t1	# E : first aligned src word complete (stall)
+	mskqh	t1, a0, t1	# U : mask leading garbage in src (stall)
+	or	t0, t1, t0	# E : first output word complete (stall)
+	or	t0, t6, t6	# E : mask original data for zero test (stall)
+
+	cmpbge	zero, t6, t8	# E :
+	beq	a2, $u_eocfin	# U :
+	lda	t6, -1		# E :
+	nop
+
+	bne	t8, $u_final	# U :
+	mskql	t6, a1, t6	# U : mask out bits already seen
+	stq_u	t0, 0(a0)	# L : store first output word
+	or      t6, t2, t2	# E : (stall)
+
+	cmpbge	zero, t2, t8	# E : find nulls in second partial
+	addq	a0, 8, a0	# E :
+	subq	a2, 1, a2	# E :
+	bne	t8, $u_late_head_exit	# U :
+
+	/* Finally, we've got all the stupid leading edge cases taken care
+	   of and we can set up to enter the main loop.  */
+	extql	t2, a1, t1	# U : position hi-bits of lo word
+	beq	a2, $u_eoc	# U :
+	ldq_u	t2, 8(a1)	# L : read next high-order source word
+	addq	a1, 8, a1	# E :
+
+	extqh	t2, a1, t0	# U : position lo-bits of hi word (stall)
+	cmpbge	zero, t2, t8	# E :
+	nop
+	bne	t8, $u_eos	# U :
+
+	/* Unaligned copy main loop.  In order to avoid reading too much,
+	   the loop is structured to detect zeros in aligned source words.
+	   This has, unfortunately, effectively pulled half of a loop
+	   iteration out into the head and half into the tail, but it does
+	   prevent nastiness from accumulating in the very thing we want
+	   to run as fast as possible.
+
+	   On entry to this basic block:
+	   t0 == the shifted low-order bits from the current source word
+	   t1 == the shifted high-order bits from the previous source word
+	   t2 == the unshifted current source word
+
+	   We further know that t2 does not contain a null terminator.  */
+
+	.align 4
+$u_loop:
+	or	t0, t1, t0	# E : current dst word now complete
+	subq	a2, 1, a2	# E : decrement word count
+	extql	t2, a1, t1	# U : extract low bits for next time
+	addq	a0, 8, a0	# E :
+
+	stq_u	t0, -8(a0)	# U : save the current word
+	beq	a2, $u_eoc	# U :
+	ldq_u	t2, 8(a1)	# U : Latency=3 load high word for next time
+	addq	a1, 8, a1	# E :
+
+	extqh	t2, a1, t0	# U : extract low bits (2 cycle stall)
+	cmpbge	zero, t2, t8	# E : test new word for eos
+	nop
+	beq	t8, $u_loop	# U :
+
+	/* We've found a zero somewhere in the source word we just read.
+	   If it resides in the lower half, we have one (probably partial)
+	   word to write out, and if it resides in the upper half, we
+	   have one full and one partial word left to write out.
+
+	   On entry to this basic block:
+	   t0 == the shifted low-order bits from the current source word
+	   t1 == the shifted high-order bits from the previous source word
+	   t2 == the unshifted current source word.  */
+$u_eos:
+	or	t0, t1, t0	# E : first (partial) source word complete
+	nop
+	cmpbge	zero, t0, t8	# E : is the null in this first bit? (stall)
+	bne	t8, $u_final	# U : (stall)
+
+	stq_u	t0, 0(a0)	# L : the null was in the high-order bits
+	addq	a0, 8, a0	# E :
+	subq	a2, 1, a2	# E :
+	nop
+
+$u_late_head_exit:
+	extql	t2, a1, t0	# U :
+	cmpbge	zero, t0, t8	# E :
+	or	t8, t10, t6	# E : (stall)
+	cmoveq	a2, t6, t8	# E : Latency=2, extra map slot (stall)
+
+	/* Take care of a final (probably partial) result word.
+	   On entry to this basic block:
+	   t0 == assembled source word
+	   t8 == cmpbge mask that found the null.  */
+$u_final:
+	negq	t8, t6		# E : isolate low bit set
+	and	t6, t8, t12	# E : (stall)
+	and	t12, 0x80, t6	# E : avoid dest word load if we can (stall)
+	bne	t6, 1f		# U : (stall)
+
+	ldq_u	t1, 0(a0)	# L :
+	subq	t12, 1, t6	# E :
+	or	t6, t12, t8	# E : (stall)
+	zapnot	t0, t8, t0	# U : kill source bytes > null
+
+	zap	t1, t8, t1	# U : kill dest bytes <= null
+	or	t0, t1, t0	# E : (stall)
+	nop
+	nop
+
+1:	stq_u	t0, 0(a0)	# L :
+	ret	(t9)		# L0 : Latency=3
+
+	  /* Got to end-of-count before end of string.  
+	     On entry to this basic block:
+	     t1 == the shifted high-order bits from the previous source word  */
+$u_eoc:
+	and	a1, 7, t6	# E : avoid final load if possible
+	sll	t10, t6, t6	# U : (stall)
+	and	t6, 0xff, t6	# E : (stall)
+	bne	t6, 1f		# U : (stall)
+
+	ldq_u	t2, 8(a1)	# L : load final src word
+	nop
+	extqh	t2, a1, t0	# U : extract low bits for last word (stall)
+	or	t1, t0, t1	# E : (stall)
+
+1:	cmpbge	zero, t1, t8	# E :
+	mov	t1, t0		# E :
+
+$u_eocfin:			# end-of-count, final word
+	or	t10, t8, t8	# E :
+	br	$u_final	# L0 : Latency=3
+
+	/* Unaligned copy entry point.  */
+	.align 4
+$unaligned:
+
+	ldq_u	t1, 0(a1)	# L : load first source word
+	and	a0, 7, t4	# E : find dest misalignment
+	and	a1, 7, t5	# E : find src misalignment
+	/* Conditionally load the first destination word and a bytemask
+	   with 0xff indicating that the destination byte is sacrosanct.  */
+	mov	zero, t0	# E :
+
+	mov	zero, t6	# E :
+	beq	t4, 1f		# U :
+	ldq_u	t0, 0(a0)	# L :
+	lda	t6, -1		# E :
+
+	mskql	t6, a0, t6	# U :
+	nop
+	nop
+	subq	a1, t4, a1	# E : sub dest misalignment from src addr
+
+	/* If source misalignment is larger than dest misalignment, we need
+	   extra startup checks to avoid SEGV.  */
+
+1:	cmplt	t4, t5, t12	# E :
+	extql	t1, a1, t1	# U : shift src into place
+	lda	t2, -1		# E : for creating masks later
+	beq	t12, $u_head	# U : (stall)
+
+	extql	t2, a1, t2	# U :
+	cmpbge	zero, t1, t8	# E : is there a zero?
+	andnot	t2, t6, t2	# E : dest mask for a single word copy
+	or	t8, t10, t5	# E : test for end-of-count too
+
+	cmpbge	zero, t2, t3	# E :
+	cmoveq	a2, t5, t8	# E : Latency=2, extra map slot
+	nop			# E : keep with cmoveq
+	andnot	t8, t3, t8	# E : (stall)
+
+	beq	t8, $u_head	# U :
+	/* At this point we've found a zero in the first partial word of
+	   the source.  We need to isolate the valid source data and mask
+	   it into the original destination data.  (Incidentally, we know
+	   that we'll need at least one byte of that original dest word.) */
+	ldq_u	t0, 0(a0)	# L :
+	negq	t8, t6		# E : build bitmask of bytes <= zero
+	mskqh	t1, t4, t1	# U :
+
+	and	t6, t8, t12	# E :
+	subq	t12, 1, t6	# E : (stall)
+	or	t6, t12, t8	# E : (stall)
+	zapnot	t2, t8, t2	# U : prepare source word; mirror changes (stall)
+
+	zapnot	t1, t8, t1	# U : to source validity mask
+	andnot	t0, t2, t0	# E : zero place for source to reside
+	or	t0, t1, t0	# E : and put it there (stall both t0, t1)
+	stq_u	t0, 0(a0)	# L : (stall)
+
+	ret	(t9)		# L0 : Latency=3
+	nop
+	nop
+	nop
+
+	.end __stxncpy