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, 711 insertions, 0 deletions

diff --git a/arch/x86/math-emu/reg_round.S b/arch/x86/math-emu/reg_round.S
new file mode 100644
index 000000000..0bb2a0921
--- /dev/null
+++ b/arch/x86/math-emu/reg_round.S
@@ -0,0 +1,711 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+	.file "reg_round.S"
+/*---------------------------------------------------------------------------+
+ |  reg_round.S                                                              |
+ |                                                                           |
+ | Rounding/truncation/etc for FPU basic arithmetic functions.               |
+ |                                                                           |
+ | Copyright (C) 1993,1995,1997                                              |
+ |                       W. Metzenthen, 22 Parker St, Ormond, Vic 3163,      |
+ |                       Australia.  E-mail billm@suburbia.net               |
+ |                                                                           |
+ | This code has four possible entry points.                                 |
+ | The following must be entered by a jmp instruction:                       |
+ |   fpu_reg_round, fpu_reg_round_sqrt, and fpu_Arith_exit.                  |
+ |                                                                           |
+ | The FPU_round entry point is intended to be used by C code.               |
+ | From C, call as:                                                          |
+ |  int FPU_round(FPU_REG *arg, unsigned int extent, unsigned int control_w) |
+ |                                                                           |
+ |    Return value is the tag of the answer, or-ed with FPU_Exception if     |
+ |    one was raised, or -1 on internal error.                               |
+ |                                                                           |
+ | For correct "up" and "down" rounding, the argument must have the correct  |
+ | sign.                                                                     |
+ |                                                                           |
+ +---------------------------------------------------------------------------*/
+
+/*---------------------------------------------------------------------------+
+ | Four entry points.                                                        |
+ |                                                                           |
+ | Needed by both the fpu_reg_round and fpu_reg_round_sqrt entry points:     |
+ |  %eax:%ebx  64 bit significand                                            |
+ |  %edx       32 bit extension of the significand                           |
+ |  %edi       pointer to an FPU_REG for the result to be stored             |
+ |  stack      calling function must have set up a C stack frame and         |
+ |             pushed %esi, %edi, and %ebx                                   |
+ |                                                                           |
+ | Needed just for the fpu_reg_round_sqrt entry point:                       |
+ |  %cx  A control word in the same format as the FPU control word.          |
+ | Otherwise, PARAM4 must give such a value.                                 |
+ |                                                                           |
+ |                                                                           |
+ | The significand and its extension are assumed to be exact in the          |
+ | following sense:                                                          |
+ |   If the significand by itself is the exact result then the significand   |
+ |   extension (%edx) must contain 0, otherwise the significand extension    |
+ |   must be non-zero.                                                       |
+ |   If the significand extension is non-zero then the significand is        |
+ |   smaller than the magnitude of the correct exact result by an amount     |
+ |   greater than zero and less than one ls bit of the significand.          |
+ |   The significand extension is only required to have three possible       |
+ |   non-zero values:                                                        |
+ |       less than 0x80000000  <=> the significand is less than 1/2 an ls    |
+ |                                 bit smaller than the magnitude of the     |
+ |                                 true exact result.                        |
+ |         exactly 0x80000000  <=> the significand is exactly 1/2 an ls bit  |
+ |                                 smaller than the magnitude of the true    |
+ |                                 exact result.                             |
+ |    greater than 0x80000000  <=> the significand is more than 1/2 an ls    |
+ |                                 bit smaller than the magnitude of the     |
+ |                                 true exact result.                        |
+ |                                                                           |
+ +---------------------------------------------------------------------------*/
+
+/*---------------------------------------------------------------------------+
+ |  The code in this module has become quite complex, but it should handle   |
+ |  all of the FPU flags which are set at this stage of the basic arithmetic |
+ |  computations.                                                            |
+ |  There are a few rare cases where the results are not set identically to  |
+ |  a real FPU. These require a bit more thought because at this stage the   |
+ |  results of the code here appear to be more consistent...                 |
+ |  This may be changed in a future version.                                 |
+ +---------------------------------------------------------------------------*/
+
+
+#include "fpu_emu.h"
+#include "exception.h"
+#include "control_w.h"
+
+/* Flags for FPU_bits_lost */
+#define	LOST_DOWN	$1
+#define	LOST_UP		$2
+
+/* Flags for FPU_denormal */
+#define	DENORMAL	$1
+#define	UNMASKED_UNDERFLOW $2
+
+
+#ifndef NON_REENTRANT_FPU
+/*	Make the code re-entrant by putting
+	local storage on the stack: */
+#define FPU_bits_lost	(%esp)
+#define FPU_denormal	1(%esp)
+
+#else
+/*	Not re-entrant, so we can gain speed by putting
+	local storage in a static area: */
+.data
+	.align 4,0
+FPU_bits_lost:
+	.byte	0
+FPU_denormal:
+	.byte	0
+#endif /* NON_REENTRANT_FPU */
+
+
+.text
+.globl fpu_reg_round
+.globl fpu_Arith_exit
+
+/* Entry point when called from C */
+SYM_FUNC_START(FPU_round)
+	pushl	%ebp
+	movl	%esp,%ebp
+	pushl	%esi
+	pushl	%edi
+	pushl	%ebx
+
+	movl	PARAM1,%edi
+	movl	SIGH(%edi),%eax
+	movl	SIGL(%edi),%ebx
+	movl	PARAM2,%edx
+
+fpu_reg_round:			/* Normal entry point */
+	movl	PARAM4,%ecx
+
+#ifndef NON_REENTRANT_FPU
+	pushl	%ebx		/* adjust the stack pointer */
+#endif /* NON_REENTRANT_FPU */ 
+
+#ifdef PARANOID
+/* Cannot use this here yet */
+/*	orl	%eax,%eax */
+/*	jns	L_entry_bugged */
+#endif /* PARANOID */
+
+	cmpw	EXP_UNDER,EXP(%edi)
+	jle	L_Make_denorm			/* The number is a de-normal */
+
+	movb	$0,FPU_denormal			/* 0 -> not a de-normal */
+
+Denorm_done:
+	movb	$0,FPU_bits_lost		/* No bits yet lost in rounding */
+
+	movl	%ecx,%esi
+	andl	CW_PC,%ecx
+	cmpl	PR_64_BITS,%ecx
+	je	LRound_To_64
+
+	cmpl	PR_53_BITS,%ecx
+	je	LRound_To_53
+
+	cmpl	PR_24_BITS,%ecx
+	je	LRound_To_24
+
+#ifdef PECULIAR_486
+/* With the precision control bits set to 01 "(reserved)", a real 80486
+   behaves as if the precision control bits were set to 11 "64 bits" */
+	cmpl	PR_RESERVED_BITS,%ecx
+	je	LRound_To_64
+#ifdef PARANOID
+	jmp	L_bugged_denorm_486
+#endif /* PARANOID */ 
+#else
+#ifdef PARANOID
+	jmp	L_bugged_denorm	/* There is no bug, just a bad control word */
+#endif /* PARANOID */ 
+#endif /* PECULIAR_486 */
+
+
+/* Round etc to 24 bit precision */
+LRound_To_24:
+	movl	%esi,%ecx
+	andl	CW_RC,%ecx
+	cmpl	RC_RND,%ecx
+	je	LRound_nearest_24
+
+	cmpl	RC_CHOP,%ecx
+	je	LCheck_truncate_24
+
+	cmpl	RC_UP,%ecx		/* Towards +infinity */
+	je	LUp_24
+
+	cmpl	RC_DOWN,%ecx		/* Towards -infinity */
+	je	LDown_24
+
+#ifdef PARANOID
+	jmp	L_bugged_round24
+#endif /* PARANOID */ 
+
+LUp_24:
+	cmpb	SIGN_POS,PARAM5
+	jne	LCheck_truncate_24	/* If negative then  up==truncate */
+
+	jmp	LCheck_24_round_up
+
+LDown_24:
+	cmpb	SIGN_POS,PARAM5
+	je	LCheck_truncate_24	/* If positive then  down==truncate */
+
+LCheck_24_round_up:
+	movl	%eax,%ecx
+	andl	$0x000000ff,%ecx
+	orl	%ebx,%ecx
+	orl	%edx,%ecx
+	jnz	LDo_24_round_up
+	jmp	L_Re_normalise
+
+LRound_nearest_24:
+	/* Do rounding of the 24th bit if needed (nearest or even) */
+	movl	%eax,%ecx
+	andl	$0x000000ff,%ecx
+	cmpl	$0x00000080,%ecx
+	jc	LCheck_truncate_24	/* less than half, no increment needed */
+
+	jne	LGreater_Half_24	/* greater than half, increment needed */
+
+	/* Possibly half, we need to check the ls bits */
+	orl	%ebx,%ebx
+	jnz	LGreater_Half_24	/* greater than half, increment needed */
+
+	orl	%edx,%edx
+	jnz	LGreater_Half_24	/* greater than half, increment needed */
+
+	/* Exactly half, increment only if 24th bit is 1 (round to even) */
+	testl	$0x00000100,%eax
+	jz	LDo_truncate_24
+
+LGreater_Half_24:			/* Rounding: increment at the 24th bit */
+LDo_24_round_up:
+	andl	$0xffffff00,%eax	/* Truncate to 24 bits */
+	xorl	%ebx,%ebx
+	movb	LOST_UP,FPU_bits_lost
+	addl	$0x00000100,%eax
+	jmp	LCheck_Round_Overflow
+
+LCheck_truncate_24:
+	movl	%eax,%ecx
+	andl	$0x000000ff,%ecx
+	orl	%ebx,%ecx
+	orl	%edx,%ecx
+	jz	L_Re_normalise		/* No truncation needed */
+
+LDo_truncate_24:
+	andl	$0xffffff00,%eax	/* Truncate to 24 bits */
+	xorl	%ebx,%ebx
+	movb	LOST_DOWN,FPU_bits_lost
+	jmp	L_Re_normalise
+
+
+/* Round etc to 53 bit precision */
+LRound_To_53:
+	movl	%esi,%ecx
+	andl	CW_RC,%ecx
+	cmpl	RC_RND,%ecx
+	je	LRound_nearest_53
+
+	cmpl	RC_CHOP,%ecx
+	je	LCheck_truncate_53
+
+	cmpl	RC_UP,%ecx		/* Towards +infinity */
+	je	LUp_53
+
+	cmpl	RC_DOWN,%ecx		/* Towards -infinity */
+	je	LDown_53
+
+#ifdef PARANOID
+	jmp	L_bugged_round53
+#endif /* PARANOID */ 
+
+LUp_53:
+	cmpb	SIGN_POS,PARAM5
+	jne	LCheck_truncate_53	/* If negative then  up==truncate */
+
+	jmp	LCheck_53_round_up
+
+LDown_53:
+	cmpb	SIGN_POS,PARAM5
+	je	LCheck_truncate_53	/* If positive then  down==truncate */
+
+LCheck_53_round_up:
+	movl	%ebx,%ecx
+	andl	$0x000007ff,%ecx
+	orl	%edx,%ecx
+	jnz	LDo_53_round_up
+	jmp	L_Re_normalise
+
+LRound_nearest_53:
+	/* Do rounding of the 53rd bit if needed (nearest or even) */
+	movl	%ebx,%ecx
+	andl	$0x000007ff,%ecx
+	cmpl	$0x00000400,%ecx
+	jc	LCheck_truncate_53	/* less than half, no increment needed */
+
+	jnz	LGreater_Half_53	/* greater than half, increment needed */
+
+	/* Possibly half, we need to check the ls bits */
+	orl	%edx,%edx
+	jnz	LGreater_Half_53	/* greater than half, increment needed */
+
+	/* Exactly half, increment only if 53rd bit is 1 (round to even) */
+	testl	$0x00000800,%ebx
+	jz	LTruncate_53
+
+LGreater_Half_53:			/* Rounding: increment at the 53rd bit */
+LDo_53_round_up:
+	movb	LOST_UP,FPU_bits_lost
+	andl	$0xfffff800,%ebx	/* Truncate to 53 bits */
+	addl	$0x00000800,%ebx
+	adcl	$0,%eax
+	jmp	LCheck_Round_Overflow
+
+LCheck_truncate_53:
+	movl	%ebx,%ecx
+	andl	$0x000007ff,%ecx
+	orl	%edx,%ecx
+	jz	L_Re_normalise
+
+LTruncate_53:
+	movb	LOST_DOWN,FPU_bits_lost
+	andl	$0xfffff800,%ebx	/* Truncate to 53 bits */
+	jmp	L_Re_normalise
+
+
+/* Round etc to 64 bit precision */
+LRound_To_64:
+	movl	%esi,%ecx
+	andl	CW_RC,%ecx
+	cmpl	RC_RND,%ecx
+	je	LRound_nearest_64
+
+	cmpl	RC_CHOP,%ecx
+	je	LCheck_truncate_64
+
+	cmpl	RC_UP,%ecx		/* Towards +infinity */
+	je	LUp_64
+
+	cmpl	RC_DOWN,%ecx		/* Towards -infinity */
+	je	LDown_64
+
+#ifdef PARANOID
+	jmp	L_bugged_round64
+#endif /* PARANOID */ 
+
+LUp_64:
+	cmpb	SIGN_POS,PARAM5
+	jne	LCheck_truncate_64	/* If negative then  up==truncate */
+
+	orl	%edx,%edx
+	jnz	LDo_64_round_up
+	jmp	L_Re_normalise
+
+LDown_64:
+	cmpb	SIGN_POS,PARAM5
+	je	LCheck_truncate_64	/* If positive then  down==truncate */
+
+	orl	%edx,%edx
+	jnz	LDo_64_round_up
+	jmp	L_Re_normalise
+
+LRound_nearest_64:
+	cmpl	$0x80000000,%edx
+	jc	LCheck_truncate_64
+
+	jne	LDo_64_round_up
+
+	/* Now test for round-to-even */
+	testb	$1,%bl
+	jz	LCheck_truncate_64
+
+LDo_64_round_up:
+	movb	LOST_UP,FPU_bits_lost
+	addl	$1,%ebx
+	adcl	$0,%eax
+
+LCheck_Round_Overflow:
+	jnc	L_Re_normalise
+
+	/* Overflow, adjust the result (significand to 1.0) */
+	rcrl	$1,%eax
+	rcrl	$1,%ebx
+	incw	EXP(%edi)
+	jmp	L_Re_normalise
+
+LCheck_truncate_64:
+	orl	%edx,%edx
+	jz	L_Re_normalise
+
+LTruncate_64:
+	movb	LOST_DOWN,FPU_bits_lost
+
+L_Re_normalise:
+	testb	$0xff,FPU_denormal
+	jnz	Normalise_result
+
+L_Normalised:
+	movl	TAG_Valid,%edx
+
+L_deNormalised:
+	cmpb	LOST_UP,FPU_bits_lost
+	je	L_precision_lost_up
+
+	cmpb	LOST_DOWN,FPU_bits_lost
+	je	L_precision_lost_down
+
+L_no_precision_loss:
+	/* store the result */
+
+L_Store_significand:
+	movl	%eax,SIGH(%edi)
+	movl	%ebx,SIGL(%edi)
+
+	cmpw	EXP_OVER,EXP(%edi)
+	jge	L_overflow
+
+	movl	%edx,%eax
+
+	/* Convert the exponent to 80x87 form. */
+	addw	EXTENDED_Ebias,EXP(%edi)
+	andw	$0x7fff,EXP(%edi)
+
+fpu_reg_round_signed_special_exit:
+
+	cmpb	SIGN_POS,PARAM5
+	je	fpu_reg_round_special_exit
+
+	orw	$0x8000,EXP(%edi)	/* Negative sign for the result. */
+
+fpu_reg_round_special_exit:
+
+#ifndef NON_REENTRANT_FPU
+	popl	%ebx		/* adjust the stack pointer */
+#endif /* NON_REENTRANT_FPU */ 
+
+fpu_Arith_exit:
+	popl	%ebx
+	popl	%edi
+	popl	%esi
+	leave
+	RET
+
+
+/*
+ * Set the FPU status flags to represent precision loss due to
+ * round-up.
+ */
+L_precision_lost_up:
+	push	%edx
+	push	%eax
+	call	set_precision_flag_up
+	popl	%eax
+	popl	%edx
+	jmp	L_no_precision_loss
+
+/*
+ * Set the FPU status flags to represent precision loss due to
+ * truncation.
+ */
+L_precision_lost_down:
+	push	%edx
+	push	%eax
+	call	set_precision_flag_down
+	popl	%eax
+	popl	%edx
+	jmp	L_no_precision_loss
+
+
+/*
+ * The number is a denormal (which might get rounded up to a normal)
+ * Shift the number right the required number of bits, which will
+ * have to be undone later...
+ */
+L_Make_denorm:
+	/* The action to be taken depends upon whether the underflow
+	   exception is masked */
+	testb	CW_Underflow,%cl		/* Underflow mask. */
+	jz	Unmasked_underflow		/* Do not make a denormal. */
+
+	movb	DENORMAL,FPU_denormal
+
+	pushl	%ecx		/* Save */
+	movw	EXP_UNDER+1,%cx
+	subw	EXP(%edi),%cx
+
+	cmpw	$64,%cx	/* shrd only works for 0..31 bits */
+	jnc	Denorm_shift_more_than_63
+
+	cmpw	$32,%cx	/* shrd only works for 0..31 bits */
+	jnc	Denorm_shift_more_than_32
+
+/*
+ * We got here without jumps by assuming that the most common requirement
+ *   is for a small de-normalising shift.
+ * Shift by [1..31] bits
+ */
+	addw	%cx,EXP(%edi)
+	orl	%edx,%edx	/* extension */
+	setne	%ch		/* Save whether %edx is non-zero */
+	xorl	%edx,%edx
+	shrd	%cl,%ebx,%edx
+	shrd	%cl,%eax,%ebx
+	shr	%cl,%eax
+	orb	%ch,%dl
+	popl	%ecx
+	jmp	Denorm_done
+
+/* Shift by [32..63] bits */
+Denorm_shift_more_than_32:
+	addw	%cx,EXP(%edi)
+	subb	$32,%cl
+	orl	%edx,%edx
+	setne	%ch
+	orb	%ch,%bl
+	xorl	%edx,%edx
+	shrd	%cl,%ebx,%edx
+	shrd	%cl,%eax,%ebx
+	shr	%cl,%eax
+	orl	%edx,%edx		/* test these 32 bits */
+	setne	%cl
+	orb	%ch,%bl
+	orb	%cl,%bl
+	movl	%ebx,%edx
+	movl	%eax,%ebx
+	xorl	%eax,%eax
+	popl	%ecx
+	jmp	Denorm_done
+
+/* Shift by [64..) bits */
+Denorm_shift_more_than_63:
+	cmpw	$64,%cx
+	jne	Denorm_shift_more_than_64
+
+/* Exactly 64 bit shift */
+	addw	%cx,EXP(%edi)
+	xorl	%ecx,%ecx
+	orl	%edx,%edx
+	setne	%cl
+	orl	%ebx,%ebx
+	setne	%ch
+	orb	%ch,%cl
+	orb	%cl,%al
+	movl	%eax,%edx
+	xorl	%eax,%eax
+	xorl	%ebx,%ebx
+	popl	%ecx
+	jmp	Denorm_done
+
+Denorm_shift_more_than_64:
+	movw	EXP_UNDER+1,EXP(%edi)
+/* This is easy, %eax must be non-zero, so.. */
+	movl	$1,%edx
+	xorl	%eax,%eax
+	xorl	%ebx,%ebx
+	popl	%ecx
+	jmp	Denorm_done
+
+
+Unmasked_underflow:
+	movb	UNMASKED_UNDERFLOW,FPU_denormal
+	jmp	Denorm_done
+
+
+/* Undo the de-normalisation. */
+Normalise_result:
+	cmpb	UNMASKED_UNDERFLOW,FPU_denormal
+	je	Signal_underflow
+
+/* The number must be a denormal if we got here. */
+#ifdef PARANOID
+	/* But check it... just in case. */
+	cmpw	EXP_UNDER+1,EXP(%edi)
+	jne	L_norm_bugged
+#endif /* PARANOID */
+
+#ifdef PECULIAR_486
+	/*
+	 * This implements a special feature of 80486 behaviour.
+	 * Underflow will be signaled even if the number is
+	 * not a denormal after rounding.
+	 * This difference occurs only for masked underflow, and not
+	 * in the unmasked case.
+	 * Actual 80486 behaviour differs from this in some circumstances.
+	 */
+	orl	%eax,%eax		/* ms bits */
+	js	LPseudoDenormal		/* Will be masked underflow */
+#else
+	orl	%eax,%eax		/* ms bits */
+	js	L_Normalised		/* No longer a denormal */
+#endif /* PECULIAR_486 */ 
+
+	jnz	LDenormal_adj_exponent
+
+	orl	%ebx,%ebx
+	jz	L_underflow_to_zero	/* The contents are zero */
+
+LDenormal_adj_exponent:
+	decw	EXP(%edi)
+
+LPseudoDenormal:
+	testb	$0xff,FPU_bits_lost	/* bits lost == underflow */
+	movl	TAG_Special,%edx
+	jz	L_deNormalised
+
+	/* There must be a masked underflow */
+	push	%eax
+	pushl	EX_Underflow
+	call	EXCEPTION
+	popl	%eax
+	popl	%eax
+	movl	TAG_Special,%edx
+	jmp	L_deNormalised
+
+
+/*
+ * The operations resulted in a number too small to represent.
+ * Masked response.
+ */
+L_underflow_to_zero:
+	push	%eax
+	call	set_precision_flag_down
+	popl	%eax
+
+	push	%eax
+	pushl	EX_Underflow
+	call	EXCEPTION
+	popl	%eax
+	popl	%eax
+
+/* Reduce the exponent to EXP_UNDER */
+	movw	EXP_UNDER,EXP(%edi)
+	movl	TAG_Zero,%edx
+	jmp	L_Store_significand
+
+
+/* The operations resulted in a number too large to represent. */
+L_overflow:
+	addw	EXTENDED_Ebias,EXP(%edi)	/* Set for unmasked response. */
+	push	%edi
+	call	arith_overflow
+	pop	%edi
+	jmp	fpu_reg_round_signed_special_exit
+
+
+Signal_underflow:
+	/* The number may have been changed to a non-denormal */
+	/* by the rounding operations. */
+	cmpw	EXP_UNDER,EXP(%edi)
+	jle	Do_unmasked_underflow
+
+	jmp	L_Normalised
+
+Do_unmasked_underflow:
+	/* Increase the exponent by the magic number */
+	addw	$(3*(1<<13)),EXP(%edi)
+	push	%eax
+	pushl	EX_Underflow
+	call	EXCEPTION
+	popl	%eax
+	popl	%eax
+	jmp	L_Normalised
+
+
+#ifdef PARANOID
+#ifdef PECULIAR_486
+L_bugged_denorm_486:
+	pushl	EX_INTERNAL|0x236
+	call	EXCEPTION
+	popl	%ebx
+	jmp	L_exception_exit
+#else
+L_bugged_denorm:
+	pushl	EX_INTERNAL|0x230
+	call	EXCEPTION
+	popl	%ebx
+	jmp	L_exception_exit
+#endif /* PECULIAR_486 */ 
+
+L_bugged_round24:
+	pushl	EX_INTERNAL|0x231
+	call	EXCEPTION
+	popl	%ebx
+	jmp	L_exception_exit
+
+L_bugged_round53:
+	pushl	EX_INTERNAL|0x232
+	call	EXCEPTION
+	popl	%ebx
+	jmp	L_exception_exit
+
+L_bugged_round64:
+	pushl	EX_INTERNAL|0x233
+	call	EXCEPTION
+	popl	%ebx
+	jmp	L_exception_exit
+
+L_norm_bugged:
+	pushl	EX_INTERNAL|0x234
+	call	EXCEPTION
+	popl	%ebx
+	jmp	L_exception_exit
+
+L_entry_bugged:
+	pushl	EX_INTERNAL|0x235
+	call	EXCEPTION
+	popl	%ebx
+L_exception_exit:
+	mov	$-1,%eax
+	jmp	fpu_reg_round_special_exit
+#endif /* PARANOID */ 
+
+SYM_FUNC_END(FPU_round)