Date: Thu, 15 Aug 2024 21:28:08 -0600 From: Warner Losh <imp@bsdimp.com> To: Zhenlei Huang <zlei@freebsd.org> Cc: Warner Losh <imp@freebsd.org>, src-committers <src-committers@freebsd.org>, "<dev-commits-src-all@freebsd.org>" <dev-commits-src-all@freebsd.org>, "<dev-commits-src-main@freebsd.org>" <dev-commits-src-main@freebsd.org> Subject: Re: git: d5507f9e4366 - main - nvme: Separate total failures from I/O failures Message-ID: <CANCZdfpk7dSiuYaU42cFXrhqMsonbPrQrjzTDeWL=vvCnoHMxA@mail.gmail.com> In-Reply-To: <0936F9D3-86A7-46AF-83E3-3EE63C41DA67@FreeBSD.org> References: <202408160230.47G2UO9J045807@gitrepo.freebsd.org> <0936F9D3-86A7-46AF-83E3-3EE63C41DA67@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
--000000000000014480061fc48d49 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable On Thu, Aug 15, 2024, 9:12=E2=80=AFPM Zhenlei Huang <zlei@freebsd.org> wrot= e: > Hi Warner, > > I'm not sure but this change seems include lots of unintended changes ( > maybe some local WIP ) . > Git am somehow screwed me. I'll fix. Warner > On Aug 16, 2024, at 10:30 AM, Warner Losh <imp@FreeBSD.org> wrote: > > > > The branch main has been updated by imp: > > > > URL: > https://cgit.FreeBSD.org/src/commit/?id=3Dd5507f9e436698ac17dc5ace7ef5849= 3988a9b04 > > > > commit d5507f9e436698ac17dc5ace7ef58493988a9b04 > > Author: Warner Losh <imp@FreeBSD.org> > > AuthorDate: 2024-08-14 22:55:49 +0000 > > Commit: Warner Losh <imp@FreeBSD.org> > > CommitDate: 2024-08-16 02:22:18 +0000 > > > > nvme: Separate total failures from I/O failures > > > > When it's a I/O failure, we can still send admin commands. Separate > out > > the admin failures and flag them as such so that we can still send > admin > > commands on half-failed drives. > > > > Fixes: 9229b3105d88 (nvme: Fail passthrough commands right away in > failed state) > > Sponsored by: Netflix > > --- > > sys/amd64/conf/IOSCHED | 2 + > > sys/amd64/conf/MPI3MR | 10 + > > sys/arm64/conf/GENERIC16K | 4 + > > .../linuxkpi/common/include/linux/#compiler.h# | 117 + > > sys/contrib/dev/iwlwifi/fw/api/soc.h | 35 + > > sys/contrib/zlib/contrib/asm686/README.686 | 51 + > > sys/contrib/zlib/contrib/asm686/match.S | 357 + > > sys/dev/ice/ice_sriov.c | 595 ++ > > sys/dev/ice/ice_sriov.h | 64 + > > sys/dev/mps/mpi/mpi2_pci.h | 141 + > > sys/dev/nvme/nvme_ctrlr.c | 46 +- > > sys/dev/nvme/nvme_private.h | 1 + > > sys/dev/nvme/nvme_qpair.c | 23 +- > > sys/dev/nvme/nvme_sim.c | 13 +- > > sys/dev/sound/pci/aureal.c | 686 ++ > > sys/dev/sound/pci/aureal.h | 99 + > > sys/dev/sound/pci/ds1-fw.h | 1602 ++++ > > sys/dev/sound/pci/ds1.c | 1103 +++ > > sys/dev/sound/pci/ds1.h | 146 + > > sys/dev/sound/pci/maestro.c | 2043 +++++ > > sys/dev/sound/pci/maestro_reg.h | 381 + > > sys/kern/bsduser-syscalls.c | 8712 > ++++++++++++++++++++ > > sys/modules/sound/driver/ds1/Makefile | 8 + > > sys/modules/sound/driver/maestro/Makefile | 8 + > > 24 files changed, 16219 insertions(+), 28 deletions(-) > > > > diff --git a/sys/amd64/conf/IOSCHED b/sys/amd64/conf/IOSCHED > > new file mode 100644 > > index 000000000000..e15106bc4c1f > > --- /dev/null > > +++ b/sys/amd64/conf/IOSCHED > > @@ -0,0 +1,2 @@ > > +include "GENERIC" > > +options CAM_IOSCHED_DYNAMIC > > diff --git a/sys/amd64/conf/MPI3MR b/sys/amd64/conf/MPI3MR > > new file mode 100644 > > index 000000000000..99e5244cb49d > > --- /dev/null > > +++ b/sys/amd64/conf/MPI3MR > > @@ -0,0 +1,10 @@ > > +include GENERIC > > + > > +device mpi3mr > > +# All the debugging options > > +options DEADLKRES # Enable the deadlock resolver > > +options INVARIANTS # Enable calls of extra sanity checking > > +options INVARIANT_SUPPORT # Extra sanity checks of internal structures= , > required by INVARIANTS > > +options QUEUE_MACRO_DEBUG_TRASH # Trash queue(2) internal pointers on > invalidation > > +options WITNESS # Enable checks to detect deadlocks and cycles > > +options WITNESS_SKIPSPIN # Don't run witness on spinlocks for speed > > diff --git a/sys/arm64/conf/GENERIC16K b/sys/arm64/conf/GENERIC16K > > new file mode 100644 > > index 000000000000..9bf9e2dadb08 > > --- /dev/null > > +++ b/sys/arm64/conf/GENERIC16K > > @@ -0,0 +1,4 @@ > > +include "GENERIC" > > + > > +ident GENERIC_16K > > + > > diff --git a/sys/compat/linuxkpi/common/include/linux/#compiler.h# > b/sys/compat/linuxkpi/common/include/linux/#compiler.h# > > new file mode 100644 > > index 000000000000..1177674aa68f > > --- /dev/null > > +++ b/sys/compat/linuxkpi/common/include/linux/#compiler.h# > > @@ -0,0 +1,117 @@ > > +/*- > > + * Copyright (c) 2010 Isilon Systems, Inc. > > + * Copyright (c) 2010 iX Systems, Inc. > > + * Copyright (c) 2010 Panasas, Inc. > > + * Copyright (c) 2013-2016 Mellanox Technologies, Ltd. > > + * Copyright (c) 2015 Fran=C3=A7ois Tigeot > > + * All rights reserved. > > + * > > + * Redistribution and use in source and binary forms, with or without > > + * modification, are permitted provided that the following conditions > > + * are met: > > + * 1. Redistributions of source code must retain the above copyright > > + * notice unmodified, this list of conditions, and the following > > + * disclaimer. > > + * 2. Redistributions in binary form must reproduce the above copyrigh= t > > + * notice, this list of conditions and the following disclaimer in > the > > + * documentation and/or other materials provided with the > distribution. > > + * > > + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS O= R > > + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED > WARRANTIES > > + * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE > DISCLAIMED. > > + * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, > > + * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING= , > BUT > > + * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS O= F > USE, > > + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON A= NY > > + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT > > + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE > USE OF > > + * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. > > + * > > + * $FreeBSD$ > > + */ > > +#ifndef _LINUX_COMPILER_H_ > > +#define _LINUX_COMPILER_H_ > > + > > +#include <sys/cdefs.h> > > + > > +#define __user > > +#define __kernel > > +#define __safe > > +#define __force > > +#define __nocast > > +#define __iomem > > +#define __chk_user_ptr(x) ((void)0) > > +#define __chk_io_ptr(x) ((void)0) > > +#define __builtin_warning(x, y...) (1) > > +#define __acquires(x) > > +#define __releases(x) > > +#define __acquire(x) do { } while (0) > > +#define __release(x) do { } while (0) > > +#define __cond_lock(x,c) (c) > > +#define __bitwise > > +#define __devinitdata > > +#define __deprecated > > +#define __init > > +#define __initconst > > +#define __devinit > > +#define __devexit > > +#define __exit > > +#define __rcu > > +#define __percpu > > +#define __weak __weak_symbol > > +#define __malloc > > +#define ___stringify(...) #__VA_ARGS__ > > +#define __stringify(...) ___stringify(__VA_ARGS__) > > +#define __attribute_const__ __attribute__((__const__)= ) > > +#undef __always_inline > > +#define __always_inline inline > > +#define noinline __noinline > > +#define ____cacheline_aligned __aligned(CACHE_LINE_SIZE= ) > > + > > +#define likely(x) __builtin_expect(!!(x), 1= ) > > +#define unlikely(x) __builtin_expect(!!(x), 0= ) > > +#define typeof(x) __typeof(x) > > + > > +#define uninitialized_var(x) x =3D x > > +#define __maybe_unused __unused > > +#define __always_unused __unused > > +#define __must_check __result_use_check > > + > > +#define __printf(a,b) __printflike(a,b) > > + > > +#define barrier() __asm__ __volatile__("": = : > :"memory") > > + > > +#if defined(LINUXKPI_VERSION) && LINUXKPI_VERSION >=3D 50000 > > +/* Moved from drm_os_freebsd.h */ > > +#define lower_32_bits(n) ((u32)(n)) > > +#define upper_32_bits(n) ((u32)(((n) >> 16) >> 16)= ) > > +#endif > > + > > +#define ___PASTE(a,b) a##b > > +#define __PASTE(a,b) ___PASTE(a,b) > > + > > +#define ACCESS_ONCE(x) (*(volatile __typeof(x) > *)&(x)) > > + > > +#define WRITE_ONCE(x,v) do { \ > > + barrier(); \ > > + ACCESS_ONCE(x) =3D (v); \ > > + barrier(); \ > > +} while (0) > > + > > +#define READ_ONCE(x) ({ \ > > + __typeof(x) __var =3D ({ \ > > + barrier(); \ > > + ACCESS_ONCE(x); \ > > + }); \ > > + barrier(); \ > > + __var; \ > > +}) > > + > > +#define lockless_dereference(p) READ_ONCE(p) > > + > > +#define _AT(T,X) ((T)(X)) > > + > > +#define __same_type(a, b) > __builtin_types_compatible_p(typeof(a), typeof(b)) > > +#define __must_be_array(a) __same_type(a, &(a)[0]) > > + > > +#endif /* _LINUX_COMPILER_H_ */ > > diff --git a/sys/contrib/dev/iwlwifi/fw/api/soc.h > b/sys/contrib/dev/iwlwifi/fw/api/soc.h > > new file mode 100644 > > index 000000000000..c5df1171462b > > --- /dev/null > > +++ b/sys/contrib/dev/iwlwifi/fw/api/soc.h > > @@ -0,0 +1,35 @@ > > +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */ > > +/* > > + * Copyright (C) 2012-2014, 2019-2020 Intel Corporation > > + * Copyright (C) 2013-2015 Intel Mobile Communications GmbH > > + * Copyright (C) 2016-2017 Intel Deutschland GmbH > > + */ > > +#ifndef __iwl_fw_api_soc_h__ > > +#define __iwl_fw_api_soc_h__ > > + > > +#define SOC_CONFIG_CMD_FLAGS_DISCRETE BIT(0) > > +#define SOC_CONFIG_CMD_FLAGS_LOW_LATENCY BIT(1) > > + > > +#define SOC_FLAGS_LTR_APPLY_DELAY_MASK 0xc > > +#define SOC_FLAGS_LTR_APPLY_DELAY_NONE 0 > > +#define SOC_FLAGS_LTR_APPLY_DELAY_200 1 > > +#define SOC_FLAGS_LTR_APPLY_DELAY_2500 2 > > +#define SOC_FLAGS_LTR_APPLY_DELAY_1820 3 > > + > > +/** > > + * struct iwl_soc_configuration_cmd - Set device stabilization latency > > + * > > + * @flags: soc settings flags. In VER_1, we can only set the DISCRETE > > + * flag, because the FW treats the whole value as an integer. In > > + * VER_2, we can set the bits independently. > > + * @latency: time for SOC to ensure stable power & XTAL > > + */ > > +struct iwl_soc_configuration_cmd { > > + __le32 flags; > > + __le32 latency; > > +} __packed; /* > > + * SOC_CONFIGURATION_CMD_S_VER_1 (see description above) > > + * SOC_CONFIGURATION_CMD_S_VER_2 > > + */ > > + > > +#endif /* __iwl_fw_api_soc_h__ */ > > diff --git a/sys/contrib/zlib/contrib/asm686/README.686 > b/sys/contrib/zlib/contrib/asm686/README.686 > > new file mode 100644 > > index 000000000000..a0bf3bea4aff > > --- /dev/null > > +++ b/sys/contrib/zlib/contrib/asm686/README.686 > > @@ -0,0 +1,51 @@ > > +This is a patched version of zlib, modified to use > > +Pentium-Pro-optimized assembly code in the deflation algorithm. The > > +files changed/added by this patch are: > > + > > +README.686 > > +match.S > > + > > +The speedup that this patch provides varies, depending on whether the > > +compiler used to build the original version of zlib falls afoul of the > > +PPro's speed traps. My own tests show a speedup of around 10-20% at > > +the default compression level, and 20-30% using -9, against a version > > +compiled using gcc 2.7.2.3. Your mileage may vary. > > + > > +Note that this code has been tailored for the PPro/PII in particular, > > +and will not perform particuarly well on a Pentium. > > + > > +If you are using an assembler other than GNU as, you will have to > > +translate match.S to use your assembler's syntax. (Have fun.) > > + > > +Brian Raiter > > +breadbox@muppetlabs.com > > +April, 1998 > > + > > + > > +Added for zlib 1.1.3: > > + > > +The patches come from > > +http://www.muppetlabs.com/~breadbox/software/assembly.html > > + > > +To compile zlib with this asm file, copy match.S to the zlib directory > > +then do: > > + > > +CFLAGS=3D"-O3 -DASMV" ./configure > > +make OBJA=3Dmatch.o > > + > > + > > +Update: > > + > > +I've been ignoring these assembly routines for years, believing that > > +gcc's generated code had caught up with it sometime around gcc 2.95 > > +and the major rearchitecting of the Pentium 4. However, I recently > > +learned that, despite what I believed, this code still has some life > > +in it. On the Pentium 4 and AMD64 chips, it continues to run about 8% > > +faster than the code produced by gcc 4.1. > > + > > +In acknowledgement of its continuing usefulness, I've altered the > > +license to match that of the rest of zlib. Share and Enjoy! > > + > > +Brian Raiter > > +breadbox@muppetlabs.com > > +April, 2007 > > diff --git a/sys/contrib/zlib/contrib/asm686/match.S > b/sys/contrib/zlib/contrib/asm686/match.S > > new file mode 100644 > > index 000000000000..fa421092785d > > --- /dev/null > > +++ b/sys/contrib/zlib/contrib/asm686/match.S > > @@ -0,0 +1,357 @@ > > +/* match.S -- x86 assembly version of the zlib longest_match() functio= n. > > + * Optimized for the Intel 686 chips (PPro and later). > > + * > > + * Copyright (C) 1998, 2007 Brian Raiter <breadbox@muppetlabs.com> > > + * > > + * This software is provided 'as-is', without any express or implied > > + * warranty. In no event will the author be held liable for any damag= es > > + * arising from the use of this software. > > + * > > + * Permission is granted to anyone to use this software for any purpos= e, > > + * including commercial applications, and to alter it and redistribute > it > > + * freely, subject to the following restrictions: > > + * > > + * 1. The origin of this software must not be misrepresented; you must > not > > + * claim that you wrote the original software. If you use this > software > > + * in a product, an acknowledgment in the product documentation > would be > > + * appreciated but is not required. > > + * 2. Altered source versions must be plainly marked as such, and must > not be > > + * misrepresented as being the original software. > > + * 3. This notice may not be removed or altered from any source > distribution. > > + */ > > + > > +#ifndef NO_UNDERLINE > > +#define match_init _match_init > > +#define longest_match _longest_match > > +#endif > > + > > +#define MAX_MATCH (258) > > +#define MIN_MATCH (3) > > +#define MIN_LOOKAHEAD (MAX_MATCH + MIN_MATCH + 1) > > +#define MAX_MATCH_8 ((MAX_MATCH + 7) & ~7) > > + > > +/* stack frame offsets */ > > + > > +#define chainlenwmask 0 /* high word: current > chain len */ > > + /* low word: s->wmask *= / > > +#define window 4 /* local copy of > s->window */ > > +#define windowbestlen 8 /* s->window + bestlen > */ > > +#define scanstart 16 /* first two bytes of > string */ > > +#define scanend 12 /* last two bytes of > string */ > > +#define scanalign 20 /* dword-misalignment of > string */ > > +#define nicematch 24 /* a good enough match > size */ > > +#define bestlen 28 /* size of best match so > far */ > > +#define scan 32 /* ptr to string wanting > match */ > > + > > +#define LocalVarsSize (36) > > +/* saved ebx 36 */ > > +/* saved edi 40 */ > > +/* saved esi 44 */ > > +/* saved ebp 48 */ > > +/* return address 52 */ > > +#define deflatestate 56 /* the function arguments > */ > > +#define curmatch 60 > > + > > +/* All the +zlib1222add offsets are due to the addition of fields > > + * in zlib in the deflate_state structure since the asm code was firs= t > written > > + * (if you compile with zlib 1.0.4 or older, use "zlib1222add equ > (-4)"). > > + * (if you compile with zlib between 1.0.5 and 1.2.2.1, use > "zlib1222add equ 0"). > > + * if you compile with zlib 1.2.2.2 or later , use "zlib1222add equ 8"= ). > > + */ > > + > > +#define zlib1222add (8) > > + > > +#define dsWSize (36+zlib1222add) > > +#define dsWMask (44+zlib1222add) > > +#define dsWindow (48+zlib1222add) > > +#define dsPrev (56+zlib1222add) > > +#define dsMatchLen (88+zlib1222add) > > +#define dsPrevMatch (92+zlib1222add) > > +#define dsStrStart (100+zlib1222add) > > +#define dsMatchStart (104+zlib1222add) > > +#define dsLookahead (108+zlib1222add) > > +#define dsPrevLen (112+zlib1222add) > > +#define dsMaxChainLen (116+zlib1222add) > > +#define dsGoodMatch (132+zlib1222add) > > +#define dsNiceMatch (136+zlib1222add) > > + > > + > > +.file "match.S" > > + > > +.globl match_init, longest_match > > + > > +.text > > + > > +/* uInt longest_match(deflate_state *deflatestate, IPos curmatch) */ > > +.cfi_sections .debug_frame > > + > > +longest_match: > > + > > +.cfi_startproc > > +/* Save registers that the compiler may be using, and adjust %esp to *= / > > +/* make room for our stack frame. *= / > > + > > + pushl %ebp > > + .cfi_def_cfa_offset 8 > > + .cfi_offset ebp, -8 > > + pushl %edi > > + .cfi_def_cfa_offset 12 > > + pushl %esi > > + .cfi_def_cfa_offset 16 > > + pushl %ebx > > + .cfi_def_cfa_offset 20 > > + subl $LocalVarsSize, %esp > > + .cfi_def_cfa_offset LocalVarsSize+20 > > + > > +/* Retrieve the function arguments. %ecx will hold cur_match *= / > > +/* throughout the entire function. %edx will hold the pointer to the *= / > > +/* deflate_state structure during the function's setup (before > */ > > +/* entering the main loop). *= / > > + > > + movl deflatestate(%esp), %edx > > + movl curmatch(%esp), %ecx > > + > > +/* uInt wmask =3D s->w_mask; = */ > > +/* unsigned chain_length =3D s->max_chain_length; > */ > > +/* if (s->prev_length >=3D s->good_match) { = */ > > +/* chain_length >>=3D 2; = */ > > +/* } *= / > > + > > + movl dsPrevLen(%edx), %eax > > + movl dsGoodMatch(%edx), %ebx > > + cmpl %ebx, %eax > > + movl dsWMask(%edx), %eax > > + movl dsMaxChainLen(%edx), %ebx > > + jl LastMatchGood > > + shrl $2, %ebx > > +LastMatchGood: > > + > > +/* chainlen is decremented once beforehand so that the function can *= / > > +/* use the sign flag instead of the zero flag for the exit test. *= / > > +/* It is then shifted into the high word, to make room for the wmask *= / > > +/* value, which it will always accompany. *= / > > + > > + decl %ebx > > + shll $16, %ebx > > + orl %eax, %ebx > > + movl %ebx, chainlenwmask(%esp) > > + > > +/* if ((uInt)nice_match > s->lookahead) nice_match =3D s->lookahead; = */ > > + > > + movl dsNiceMatch(%edx), %eax > > + movl dsLookahead(%edx), %ebx > > + cmpl %eax, %ebx > > + jl LookaheadLess > > + movl %eax, %ebx > > +LookaheadLess: movl %ebx, nicematch(%esp) > > + > > +/* register Bytef *scan =3D s->window + s->strstart; = */ > > + > > + movl dsWindow(%edx), %esi > > + movl %esi, window(%esp) > > + movl dsStrStart(%edx), %ebp > > + lea (%esi,%ebp), %edi > > + movl %edi, scan(%esp) > > + > > +/* Determine how many bytes the scan ptr is off from being *= / > > +/* dword-aligned. *= / > > + > > + movl %edi, %eax > > + negl %eax > > + andl $3, %eax > > + movl %eax, scanalign(%esp) > > + > > +/* IPos limit =3D s->strstart > (IPos)MAX_DIST(s) ? = */ > > +/* s->strstart - (IPos)MAX_DIST(s) : NIL; > */ > > + > > + movl dsWSize(%edx), %eax > > + subl $MIN_LOOKAHEAD, %eax > > + subl %eax, %ebp > > + jg LimitPositive > > + xorl %ebp, %ebp > > +LimitPositive: > > + > > +/* int best_len =3D s->prev_length; = */ > > + > > + movl dsPrevLen(%edx), %eax > > + movl %eax, bestlen(%esp) > > + > > +/* Store the sum of s->window + best_len in %esi locally, and in %esi. > */ > > + > > + addl %eax, %esi > > + movl %esi, windowbestlen(%esp) > > + > > +/* register ush scan_start =3D *(ushf*)scan; = */ > > +/* register ush scan_end =3D *(ushf*)(scan+best_len-1); > */ > > +/* Posf *prev =3D s->prev; = */ > > + > > + movzwl (%edi), %ebx > > + movl %ebx, scanstart(%esp) > > + movzwl -1(%edi,%eax), %ebx > > + movl %ebx, scanend(%esp) > > + movl dsPrev(%edx), %edi > > + > > +/* Jump into the main loop. *= / > > + > > + movl chainlenwmask(%esp), %edx > > + jmp LoopEntry > > + > > +.balign 16 > > + > > +/* do { > > + * match =3D s->window + cur_match; > > + * if (*(ushf*)(match+best_len-1) !=3D scan_end || > > + * *(ushf*)match !=3D scan_start) continue; > > + * [...] > > + * } while ((cur_match =3D prev[cur_match & wmask]) > limit > > + * && --chain_length !=3D 0); > > + * > > + * Here is the inner loop of the function. The function will spend the > > + * majority of its time in this loop, and majority of that time will > > + * be spent in the first ten instructions. > > + * > > + * Within this loop: > > + * %ebx =3D scanend > > + * %ecx =3D curmatch > > + * %edx =3D chainlenwmask - i.e., ((chainlen << 16) | wmask) > > + * %esi =3D windowbestlen - i.e., (window + bestlen) > > + * %edi =3D prev > > + * %ebp =3D limit > > + */ > > +LookupLoop: > > + andl %edx, %ecx > > + movzwl (%edi,%ecx,2), %ecx > > + cmpl %ebp, %ecx > > + jbe LeaveNow > > + subl $0x00010000, %edx > > + js LeaveNow > > +LoopEntry: movzwl -1(%esi,%ecx), %eax > > + cmpl %ebx, %eax > > + jnz LookupLoop > > + movl window(%esp), %eax > > + movzwl (%eax,%ecx), %eax > > + cmpl scanstart(%esp), %eax > > + jnz LookupLoop > > + > > +/* Store the current value of chainlen. > */ > > + > > + movl %edx, chainlenwmask(%esp) > > + > > +/* Point %edi to the string under scrutiny, and %esi to the string we > */ > > +/* are hoping to match it up with. In actuality, %esi and %edi are *= / > > +/* both pointed (MAX_MATCH_8 - scanalign) bytes ahead, and %edx is *= / > > +/* initialized to -(MAX_MATCH_8 - scanalign). > */ > > + > > + movl window(%esp), %esi > > + movl scan(%esp), %edi > > + addl %ecx, %esi > > + movl scanalign(%esp), %eax > > + movl $(-MAX_MATCH_8), %edx > > + lea MAX_MATCH_8(%edi,%eax), %edi > > + lea MAX_MATCH_8(%esi,%eax), %esi > > + > > +/* Test the strings for equality, 8 bytes at a time. At the end, > > + * adjust %edx so that it is offset to the exact byte that mismatched. > > + * > > + * We already know at this point that the first three bytes of the > > + * strings match each other, and they can be safely passed over before > > + * starting the compare loop. So what this code does is skip over 0-3 > > + * bytes, as much as necessary in order to dword-align the %edi > > + * pointer. (%esi will still be misaligned three times out of four.) > > + * > > + * It should be confessed that this loop usually does not represent > > + * much of the total running time. Replacing it with a more > > + * straightforward "rep cmpsb" would not drastically degrade > > + * performance. > > + */ > > +LoopCmps: > > + movl (%esi,%edx), %eax > > + xorl (%edi,%edx), %eax > > + jnz LeaveLoopCmps > > + movl 4(%esi,%edx), %eax > > + xorl 4(%edi,%edx), %eax > > + jnz LeaveLoopCmps4 > > + addl $8, %edx > > + jnz LoopCmps > > + jmp LenMaximum > > +LeaveLoopCmps4: addl $4, %edx > > +LeaveLoopCmps: testl $0x0000FFFF, %eax > > + jnz LenLower > > + addl $2, %edx > > + shrl $16, %eax > > +LenLower: subb $1, %al > > + adcl $0, %edx > > + > > +/* Calculate the length of the match. If it is longer than MAX_MATCH, > */ > > +/* then automatically accept it as the best possible match and leave. > */ > > + > > + lea (%edi,%edx), %eax > > + movl scan(%esp), %edi > > + subl %edi, %eax > > + cmpl $MAX_MATCH, %eax > > + jge LenMaximum > > + > > +/* If the length of the match is not longer than the best match we *= / > > +/* have so far, then forget it and return to the lookup loop. > */ > > + > > + movl deflatestate(%esp), %edx > > + movl bestlen(%esp), %ebx > > + cmpl %ebx, %eax > > + jg LongerMatch > > + movl windowbestlen(%esp), %esi > > + movl dsPrev(%edx), %edi > > + movl scanend(%esp), %ebx > > + movl chainlenwmask(%esp), %edx > > + jmp LookupLoop > > + > > +/* s->match_start =3D cur_match; > */ > > +/* best_len =3D len; = */ > > +/* if (len >=3D nice_match) break; = */ > > +/* scan_end =3D *(ushf*)(scan+best_len-1); = */ > > + > > +LongerMatch: movl nicematch(%esp), %ebx > > + movl %eax, bestlen(%esp) > > + movl %ecx, dsMatchStart(%edx) > > + cmpl %ebx, %eax > > + jge LeaveNow > > + movl window(%esp), %esi > > + addl %eax, %esi > > + movl %esi, windowbestlen(%esp) > > + movzwl -1(%edi,%eax), %ebx > > + movl dsPrev(%edx), %edi > > + movl %ebx, scanend(%esp) > > + movl chainlenwmask(%esp), %edx > > + jmp LookupLoop > > + > > +/* Accept the current string, with the maximum possible length. > */ > > + > > +LenMaximum: movl deflatestate(%esp), %edx > > + movl $MAX_MATCH, bestlen(%esp) > > + movl %ecx, dsMatchStart(%edx) > > + > > +/* if ((uInt)best_len <=3D s->lookahead) return (uInt)best_len; > */ > > +/* return s->lookahead; > */ > > + > > +LeaveNow: > > + movl deflatestate(%esp), %edx > > + movl bestlen(%esp), %ebx > > + movl dsLookahead(%edx), %eax > > + cmpl %eax, %ebx > > + jg LookaheadRet > > + movl %ebx, %eax > > +LookaheadRet: > > + > > +/* Restore the stack and return from whence we came. *= / > > + > > + addl $LocalVarsSize, %esp > > + .cfi_def_cfa_offset 20 > > + popl %ebx > > + .cfi_def_cfa_offset 16 > > + popl %esi > > + .cfi_def_cfa_offset 12 > > + popl %edi > > + .cfi_def_cfa_offset 8 > > + popl %ebp > > + .cfi_def_cfa_offset 4 > > +.cfi_endproc > > +match_init: ret > > diff --git a/sys/dev/ice/ice_sriov.c b/sys/dev/ice/ice_sriov.c > > new file mode 100644 > > index 000000000000..c0521e667fa2 > > --- /dev/null > > +++ b/sys/dev/ice/ice_sriov.c > > @@ -0,0 +1,595 @@ > > +/* SPDX-License-Identifier: BSD-3-Clause */ > > +/* Copyright (c) 2021, Intel Corporation > > + * All rights reserved. > > + * > > + * Redistribution and use in source and binary forms, with or without > > + * modification, are permitted provided that the following conditions > are met: > > + * > > + * 1. Redistributions of source code must retain the above copyright > notice, > > + * this list of conditions and the following disclaimer. > > + * > > + * 2. Redistributions in binary form must reproduce the above > copyright > > + * notice, this list of conditions and the following disclaimer i= n > the > > + * documentation and/or other materials provided with the > distribution. > > + * > > + * 3. Neither the name of the Intel Corporation nor the names of its > > + * contributors may be used to endorse or promote products derive= d > from > > + * this software without specific prior written permission. > > + * > > + * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTOR= S > "AS IS" > > + * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED > TO, THE > > + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR > PURPOSE > > + * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR > CONTRIBUTORS BE > > + * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, O= R > > + * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT = OF > > + * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR > BUSINESS > > + * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, > WHETHER IN > > + * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR > OTHERWISE) > > + * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISE= D > OF THE > > + * POSSIBILITY OF SUCH DAMAGE. > > + */ > > + > > +#include "ice_common.h" > > +#include "ice_sriov.h" > > + > > +/** > > + * ice_aq_send_msg_to_vf > > + * @hw: pointer to the hardware structure > > + * @vfid: VF ID to send msg > > + * @v_opcode: opcodes for VF-PF communication > > + * @v_retval: return error code > > + * @msg: pointer to the msg buffer > > + * @msglen: msg length > > + * @cd: pointer to command details > > + * > > + * Send message to VF driver (0x0802) using mailbox > > + * queue and asynchronously sending message via > > + * ice_sq_send_cmd() function > > + */ > > +enum ice_status > > +ice_aq_send_msg_to_vf(struct ice_hw *hw, u16 vfid, u32 v_opcode, u32 > v_retval, > > + u8 *msg, u16 msglen, struct ice_sq_cd *cd) > > +{ > > + struct ice_aqc_pf_vf_msg *cmd; > > + struct ice_aq_desc desc; > > + > > + ice_fill_dflt_direct_cmd_desc(&desc, ice_mbx_opc_send_msg_to_vf); > > + > > + cmd =3D &desc.params.virt; > > + cmd->id =3D CPU_TO_LE32(vfid); > > + > > + desc.cookie_high =3D CPU_TO_LE32(v_opcode); > > + desc.cookie_low =3D CPU_TO_LE32(v_retval); > > + > > + if (msglen) > > + desc.flags |=3D CPU_TO_LE16(ICE_AQ_FLAG_RD); > > + > > + return ice_sq_send_cmd(hw, &hw->mailboxq, &desc, msg, msglen, cd)= ; > > +} > > + > > +/** > > + * ice_aq_send_msg_to_pf > > + * @hw: pointer to the hardware structure > > + * @v_opcode: opcodes for VF-PF communication > > + * @v_retval: return error code > > + * @msg: pointer to the msg buffer > > + * @msglen: msg length > > + * @cd: pointer to command details > > + * > > + * Send message to PF driver using mailbox queue. By default, this > > + * message is sent asynchronously, i.e. ice_sq_send_cmd() > > + * does not wait for completion before returning. > > + */ > > +enum ice_status > > +ice_aq_send_msg_to_pf(struct ice_hw *hw, enum virtchnl_ops v_opcode, > > + enum ice_status v_retval, u8 *msg, u16 msglen, > > + struct ice_sq_cd *cd) > > +{ > > + struct ice_aq_desc desc; > > + > > + ice_fill_dflt_direct_cmd_desc(&desc, ice_mbx_opc_send_msg_to_pf); > > + desc.cookie_high =3D CPU_TO_LE32(v_opcode); > > + desc.cookie_low =3D CPU_TO_LE32(v_retval); > > + > > + if (msglen) > > + desc.flags |=3D CPU_TO_LE16(ICE_AQ_FLAG_RD); > > + > > + return ice_sq_send_cmd(hw, &hw->mailboxq, &desc, msg, msglen, cd)= ; > > +} > > + > > +/** > > + * ice_conv_link_speed_to_virtchnl > > + * @adv_link_support: determines the format of the returned link speed > > + * @link_speed: variable containing the link_speed to be converted > > + * > > + * Convert link speed supported by HW to link speed supported by > virtchnl. > > + * If adv_link_support is true, then return link speed in Mbps. Else > return > > + * link speed as a VIRTCHNL_LINK_SPEED_* casted to a u32. Note that th= e > caller > > + * needs to cast back to an enum virtchnl_link_speed in the case where > > + * adv_link_support is false, but when adv_link_support is true the > caller can > > + * expect the speed in Mbps. > > + */ > > +u32 ice_conv_link_speed_to_virtchnl(bool adv_link_support, u16 > link_speed) > > +{ > > + u32 speed; > > + > > + if (adv_link_support) > > + switch (link_speed) { > > + case ICE_AQ_LINK_SPEED_10MB: > > + speed =3D ICE_LINK_SPEED_10MBPS; > > + break; > > + case ICE_AQ_LINK_SPEED_100MB: > > + speed =3D ICE_LINK_SPEED_100MBPS; > > + break; > > + case ICE_AQ_LINK_SPEED_1000MB: > > + speed =3D ICE_LINK_SPEED_1000MBPS; > > + break; > > + case ICE_AQ_LINK_SPEED_2500MB: > > + speed =3D ICE_LINK_SPEED_2500MBPS; > > + break; > > + case ICE_AQ_LINK_SPEED_5GB: > > + speed =3D ICE_LINK_SPEED_5000MBPS; > > + break; > > + case ICE_AQ_LINK_SPEED_10GB: > > + speed =3D ICE_LINK_SPEED_10000MBPS; > > + break; > > + case ICE_AQ_LINK_SPEED_20GB: > > + speed =3D ICE_LINK_SPEED_20000MBPS; > > + break; > > + case ICE_AQ_LINK_SPEED_25GB: > > + speed =3D ICE_LINK_SPEED_25000MBPS; > > + break; > > + case ICE_AQ_LINK_SPEED_40GB: > > + speed =3D ICE_LINK_SPEED_40000MBPS; > > + break; > > + case ICE_AQ_LINK_SPEED_50GB: > > + speed =3D ICE_LINK_SPEED_50000MBPS; > > + break; > > + case ICE_AQ_LINK_SPEED_100GB: > > + speed =3D ICE_LINK_SPEED_100000MBPS; > > + break; > > + default: > > + speed =3D ICE_LINK_SPEED_UNKNOWN; > > + break; > > + } > > + else > > + /* Virtchnl speeds are not defined for every speed > supported in > > + * the hardware. To maintain compatibility with older AVF > > + * drivers, while reporting the speed the new speed value= s > are > > + * resolved to the closest known virtchnl speeds > > + */ > > + switch (link_speed) { > > + case ICE_AQ_LINK_SPEED_10MB: > > + case ICE_AQ_LINK_SPEED_100MB: > > + speed =3D (u32)VIRTCHNL_LINK_SPEED_100MB; > > + break; > > + case ICE_AQ_LINK_SPEED_1000MB: > > + case ICE_AQ_LINK_SPEED_2500MB: > > + case ICE_AQ_LINK_SPEED_5GB: > > + speed =3D (u32)VIRTCHNL_LINK_SPEED_1GB; > > + break; > > + case ICE_AQ_LINK_SPEED_10GB: > > + speed =3D (u32)VIRTCHNL_LINK_SPEED_10GB; > > + break; > > + case ICE_AQ_LINK_SPEED_20GB: > > + speed =3D (u32)VIRTCHNL_LINK_SPEED_20GB; > > + break; > > + case ICE_AQ_LINK_SPEED_25GB: > > + speed =3D (u32)VIRTCHNL_LINK_SPEED_25GB; > > + break; > > + case ICE_AQ_LINK_SPEED_40GB: > > + case ICE_AQ_LINK_SPEED_50GB: > > + case ICE_AQ_LINK_SPEED_100GB: > > + speed =3D (u32)VIRTCHNL_LINK_SPEED_40GB; > > + break; > > + default: > > + speed =3D (u32)VIRTCHNL_LINK_SPEED_UNKNOWN; > > + break; > > + } > > + > > + return speed; > > +} > > + > > +/* The mailbox overflow detection algorithm helps to check if there > > + * is a possibility of a malicious VF transmitting too many MBX > messages to the > > + * PF. > > + * 1. The mailbox snapshot structure, ice_mbx_snapshot, is initialized > during > > + * driver initialization in ice_init_hw() using ice_mbx_init_snapshot(= ). > > + * The struct ice_mbx_snapshot helps to track and traverse a static > window of > > + * messages within the mailbox queue while looking for a malicious VF. > > + * > > + * 2. When the caller starts processing its mailbox queue in response > to an > > + * interrupt, the structure ice_mbx_snapshot is expected to be cleared > before > > + * the algorithm can be run for the first time for that interrupt. Thi= s > can be > > + * done via ice_mbx_reset_snapshot(). > > + * > > + * 3. For every message read by the caller from the MBX Queue, the > caller must > > + * call the detection algorithm's entry function > ice_mbx_vf_state_handler(). > > + * Before every call to ice_mbx_vf_state_handler() the struct > ice_mbx_data is > > + * filled as it is required to be passed to the algorithm. > > + * > > + * 4. Every time a message is read from the MBX queue, a VFId is > received which > > + * is passed to the state handler. The boolean output is_malvf of the > state > > + * handler ice_mbx_vf_state_handler() serves as an indicator to the > caller > > + * whether this VF is malicious or not. > > + * > > + * 5. When a VF is identified to be malicious, the caller can send a > message > > + * to the system administrator. The caller can invoke > ice_mbx_report_malvf() > > + * to help determine if a malicious VF is to be reported or not. This > function > > + * requires the caller to maintain a global bitmap to track all > malicious VFs > > + * and pass that to ice_mbx_report_malvf() along with the VFID which > was identified > > + * to be malicious by ice_mbx_vf_state_handler(). > > + * > > + * 6. The global bitmap maintained by PF can be cleared completely if > PF is in > > + * reset or the bit corresponding to a VF can be cleared if that VF is > in reset. > > + * When a VF is shut down and brought back up, we assume that the new = VF > > + * brought up is not malicious and hence report it if found malicious. > > + * > > + * 7. The function ice_mbx_reset_snapshot() is called to reset the > information > > + * in ice_mbx_snapshot for every new mailbox interrupt handled. > > + * > > + * 8. The memory allocated for variables in ice_mbx_snapshot is > de-allocated > > + * when driver is unloaded. > > + */ > > +#define ICE_RQ_DATA_MASK(rq_data) ((rq_data) & PF_MBX_ARQH_ARQH_M) > > +/* Using the highest value for an unsigned 16-bit value 0xFFFF to > indicate that > > + * the max messages check must be ignored in the algorithm > > + */ > > +#define ICE_IGNORE_MAX_MSG_CNT 0xFFFF > > + > > +/** > > + * ice_mbx_traverse - Pass through mailbox snapshot > > + * @hw: pointer to the HW struct > > + * @new_state: new algorithm state > > + * > > + * Traversing the mailbox static snapshot without checking > > + * for malicious VFs. > > + */ > > +static void > > +ice_mbx_traverse(struct ice_hw *hw, > > + enum ice_mbx_snapshot_state *new_state) > > +{ > > + struct ice_mbx_snap_buffer_data *snap_buf; > > + u32 num_iterations; > > + > > + snap_buf =3D &hw->mbx_snapshot.mbx_buf; > > + > > + /* As mailbox buffer is circular, applying a mask > > + * on the incremented iteration count. > > + */ > > + num_iterations =3D ICE_RQ_DATA_MASK(++snap_buf->num_iterations); > > + > > + /* Checking either of the below conditions to exit snapshot > traversal: > > + * Condition-1: If the number of iterations in the mailbox is > equal to > > + * the mailbox head which would indicate that we have reached the > end > > + * of the static snapshot. > > + * Condition-2: If the maximum messages serviced in the mailbox > for a > > + * given interrupt is the highest possible value then there is no > need > > + * to check if the number of messages processed is equal to it. I= f > not > > + * check if the number of messages processed is greater than or > equal > > + * to the maximum number of mailbox entries serviced in current > work item. > > + */ > > + if (num_iterations =3D=3D snap_buf->head || > > + (snap_buf->max_num_msgs_mbx < ICE_IGNORE_MAX_MSG_CNT && > > + ++snap_buf->num_msg_proc >=3D snap_buf->max_num_msgs_mbx)) > > + *new_state =3D ICE_MAL_VF_DETECT_STATE_NEW_SNAPSHOT; > > +} > > + > > +/** > > + * ice_mbx_detect_malvf - Detect malicious VF in snapshot > > + * @hw: pointer to the HW struct > > + * @vf_id: relative virtual function ID > > + * @new_state: new algorithm state > > + * @is_malvf: boolean output to indicate if VF is malicious > > + * > > + * This function tracks the number of asynchronous messages > > + * sent per VF and marks the VF as malicious if it exceeds > > + * the permissible number of messages to send. > > + */ > > +static enum ice_status > > +ice_mbx_detect_malvf(struct ice_hw *hw, u16 vf_id, > > + enum ice_mbx_snapshot_state *new_state, > > + bool *is_malvf) > > +{ > > + struct ice_mbx_snapshot *snap =3D &hw->mbx_snapshot; > > + > > + if (vf_id >=3D snap->mbx_vf.vfcntr_len) > > + return ICE_ERR_OUT_OF_RANGE; > > + > > + /* increment the message count in the VF array */ > > + snap->mbx_vf.vf_cntr[vf_id]++; > > + > > + if (snap->mbx_vf.vf_cntr[vf_id] >=3D ICE_ASYNC_VF_MSG_THRESHOLD) > > + *is_malvf =3D true; > > + > > + /* continue to iterate through the mailbox snapshot */ > > + ice_mbx_traverse(hw, new_state); > > + > > + return ICE_SUCCESS; > > +} > > + > > +/** > > + * ice_mbx_reset_snapshot - Reset mailbox snapshot structure > > + * @snap: pointer to mailbox snapshot structure in the ice_hw struct > > + * > > + * Reset the mailbox snapshot structure and clear VF counter array. > > + */ > > +static void ice_mbx_reset_snapshot(struct ice_mbx_snapshot *snap) > > *** 15611 LINES SKIPPED *** > > > > --000000000000014480061fc48d49 Content-Type: text/html; charset="UTF-8" Content-Transfer-Encoding: quoted-printable <div dir=3D"auto"><div><br><br><div class=3D"gmail_quote"><div dir=3D"ltr" = class=3D"gmail_attr">On Thu, Aug 15, 2024, 9:12=E2=80=AFPM Zhenlei Huang &l= t;<a href=3D"mailto:zlei@freebsd.org">zlei@freebsd.org</a>> wrote:<br></= div><blockquote class=3D"gmail_quote" style=3D"margin:0 0 0 .8ex;border-lef= t:1px #ccc solid;padding-left:1ex">Hi Warner,<br> <br> I'm not sure but this change seems include lots of unintended changes (= maybe some local WIP ) .<br></blockquote></div></div><div dir=3D"auto"><br= ></div><div dir=3D"auto">Git am somehow screwed me.=C2=A0 I'll fix.</di= v><div dir=3D"auto"><br></div><div dir=3D"auto">Warner</div><div dir=3D"aut= o"><br></div><div dir=3D"auto"><div class=3D"gmail_quote"><blockquote class= =3D"gmail_quote" style=3D"margin:0 0 0 .8ex;border-left:1px #ccc solid;padd= ing-left:1ex"> > On Aug 16, 2024, at 10:30 AM, Warner Losh <imp@FreeBSD.org> wrot= e:<br> > <br> > The branch main has been updated by imp:<br> > <br> > URL: <a href=3D"https://cgit.FreeBSD.org/src/commit/?id=3Dd5507f9e4366= 98ac17dc5ace7ef58493988a9b04" rel=3D"noreferrer noreferrer" target=3D"_blan= k">https://cgit.FreeBSD.org/src/commit/?id=3Dd5507f9e436698ac17dc5ace7ef584= 93988a9b04</a><br> > <br> > commit d5507f9e436698ac17dc5ace7ef58493988a9b04<br> > Author:=C2=A0 =C2=A0 =C2=A0Warner Losh <imp@FreeBSD.org><br> > AuthorDate: 2024-08-14 22:55:49 +0000<br> > Commit:=C2=A0 =C2=A0 =C2=A0Warner Losh <imp@FreeBSD.org><br> > CommitDate: 2024-08-16 02:22:18 +0000<br> > <br> >=C2=A0 =C2=A0 nvme: Separate total failures from I/O failures<br> > <br> >=C2=A0 =C2=A0 When it's a I/O failure, we can still send admin comm= ands. Separate out<br> >=C2=A0 =C2=A0 the admin failures and flag them as such so that we can s= till send admin<br> >=C2=A0 =C2=A0 commands on half-failed drives.<br> > <br> >=C2=A0 =C2=A0 Fixes: 9229b3105d88 (nvme: Fail passthrough commands righ= t away in failed state)<br> >=C2=A0 =C2=A0 Sponsored by: Netflix<br> > ---<br> > sys/amd64/conf/IOSCHED=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0|=C2=A0 =C2=A0 2 +<= br> > sys/amd64/conf/MPI3MR=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 |=C2=A0 =C2=A010 +<= br> > sys/arm64/conf/GENERIC16K=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 |=C2=A0 =C2=A0 4 +<br> > .../linuxkpi/common/include/linux/#compiler.h#=C2=A0 =C2=A0 =C2=A0|=C2= =A0 117 +<br> > sys/contrib/dev/iwlwifi/fw/api/soc.h=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0|=C2=A0 =C2=A035 +<br> > sys/contrib/zlib/contrib/asm686/README.686=C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0|=C2=A0 =C2=A051 +<br> > sys/contrib/zlib/contrib/asm686/match.S=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 |=C2=A0 357 +<br> > sys/dev/ice/ice_sriov.c=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 |=C2=A0 595 ++<br> > sys/dev/ice/ice_sriov.h=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 |=C2=A0 =C2=A064 +<br> > sys/dev/mps/mpi/mpi2_pci.h=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0|=C2=A0 141 +<br> > sys/dev/nvme/nvme_ctrlr.c=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 |=C2=A0 =C2=A046 +-<br> > sys/dev/nvme/nvme_private.h=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 |=C2=A0 =C2=A0 1 +<br> > sys/dev/nvme/nvme_qpair.c=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 |=C2=A0 =C2=A023 +-<br> > sys/dev/nvme/nvme_sim.c=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 |=C2=A0 =C2=A013 +-<br= > > sys/dev/sound/pci/aureal.c=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0|=C2=A0 686 ++<br> > sys/dev/sound/pci/aureal.h=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0|=C2=A0 =C2=A099 +<br> > sys/dev/sound/pci/ds1-fw.h=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0| 1602 ++++<br> > sys/dev/sound/pci/ds1.c=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 | 1103 +++<br> > sys/dev/sound/pci/ds1.h=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 |=C2=A0 146 +<br> > sys/dev/sound/pci/maestro.c=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 | 2043 +++++<br> > sys/dev/sound/pci/maestro_reg.h=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 |=C2=A0 381 +<br> > sys/kern/bsduser-syscalls.c=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 | 8712 ++++++++++++++++++++<br> > sys/modules/sound/driver/ds1/Makefile=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 |=C2=A0 =C2=A0 8 +<br> > sys/modules/sound/driver/maestro/Makefile=C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 |=C2=A0 =C2=A0 8 +<br> > 24 files changed, 16219 insertions(+), 28 deletions(-)<br> > <br> > diff --git a/sys/amd64/conf/IOSCHED b/sys/amd64/conf/IOSCHED<br> > new file mode 100644<br> > index 000000000000..e15106bc4c1f<br> > --- /dev/null<br> > +++ b/sys/amd64/conf/IOSCHED<br> > @@ -0,0 +1,2 @@<br> > +include "GENERIC"<br> > +options CAM_IOSCHED_DYNAMIC<br> > diff --git a/sys/amd64/conf/MPI3MR b/sys/amd64/conf/MPI3MR<br> > new file mode 100644<br> > index 000000000000..99e5244cb49d<br> > --- /dev/null<br> > +++ b/sys/amd64/conf/MPI3MR<br> > @@ -0,0 +1,10 @@<br> > +include GENERIC<br> > +<br> > +device mpi3mr<br> > +# All the debugging options<br> > +options DEADLKRES # Enable the deadlock resolver<br> > +options INVARIANTS # Enable calls of extra sanity checking<br> > +options INVARIANT_SUPPORT # Extra sanity checks of internal structure= s, required by INVARIANTS<br> > +options QUEUE_MACRO_DEBUG_TRASH # Trash queue(2) internal pointers on= invalidation<br> > +options WITNESS # Enable checks to detect deadlocks and cycles<br> > +options WITNESS_SKIPSPIN # Don't run witness on spinlocks for spe= ed<br> > diff --git a/sys/arm64/conf/GENERIC16K b/sys/arm64/conf/GENERIC16K<br> > new file mode 100644<br> > index 000000000000..9bf9e2dadb08<br> > --- /dev/null<br> > +++ b/sys/arm64/conf/GENERIC16K<br> > @@ -0,0 +1,4 @@<br> > +include=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 "GENERIC= "<br> > +<br> > +ident=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 GENERIC_= 16K<br> > +<br> > diff --git a/sys/compat/linuxkpi/common/include/linux/#compiler.h# b/s= ys/compat/linuxkpi/common/include/linux/#compiler.h#<br> > new file mode 100644<br> > index 000000000000..1177674aa68f<br> > --- /dev/null<br> > +++ b/sys/compat/linuxkpi/common/include/linux/#compiler.h#<br> > @@ -0,0 +1,117 @@<br> > +/*-<br> > + * Copyright (c) 2010 Isilon Systems, Inc.<br> > + * Copyright (c) 2010 iX Systems, Inc.<br> > + * Copyright (c) 2010 Panasas, Inc.<br> > + * Copyright (c) 2013-2016 Mellanox Technologies, Ltd.<br> > + * Copyright (c) 2015 Fran=C3=A7ois Tigeot<br> > + * All rights reserved.<br> > + *<br> > + * Redistribution and use in source and binary forms, with or without= <br> > + * modification, are permitted provided that the following conditions= <br> > + * are met:<br> > + * 1. Redistributions of source code must retain the above copyright<= br> > + *=C2=A0 =C2=A0 notice unmodified, this list of conditions, and the f= ollowing<br> > + *=C2=A0 =C2=A0 disclaimer.<br> > + * 2. Redistributions in binary form must reproduce the above copyrig= ht<br> > + *=C2=A0 =C2=A0 notice, this list of conditions and the following dis= claimer in the<br> > + *=C2=A0 =C2=A0 documentation and/or other materials provided with th= e distribution.<br> > + *<br> > + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY = EXPRESS OR<br> > + * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WAR= RANTIES<br> > + * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLA= IMED.<br> > + * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,<b= r> > + * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDIN= G, BUT<br> > + * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS = OF USE,<br> > + * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON = ANY<br> > + * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TOR= T<br> > + * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE = USE OF<br> > + * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.<= br> > + *<br> > + * $FreeBSD$<br> > + */<br> > +#ifndef=C2=A0 =C2=A0 =C2=A0 _LINUX_COMPILER_H_<br> > +#define=C2=A0 =C2=A0 =C2=A0 _LINUX_COMPILER_H_<br> > +<br> > +#include <sys/cdefs.h><br> > +<br> > +#define __user<br> > +#define __kernel<br> > +#define __safe<br> > +#define __force<br> > +#define __nocast<br> > +#define __iomem<br> > +#define __chk_user_ptr(x)=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 ((= void)0)<br> > +#define __chk_io_ptr(x)=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 ((void)0)<br> > +#define __builtin_warning(x, y...)=C2=A0 =C2=A0(1)<br> > +#define __acquires(x)<br> > +#define __releases(x)<br> > +#define __acquire(x)=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0do { } while (0)<br> > +#define __release(x)=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0do { } while (0)<br> > +#define __cond_lock(x,c)=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0(c)<br> > +#define=C2=A0 =C2=A0 =C2=A0 __bitwise<br> > +#define __devinitdata<br> > +#define=C2=A0 =C2=A0 =C2=A0 __deprecated<br> > +#define __init<br> > +#define=C2=A0 =C2=A0 =C2=A0 __initconst<br> > +#define=C2=A0 =C2=A0 =C2=A0 __devinit<br> > +#define=C2=A0 =C2=A0 =C2=A0 __devexit<br> > +#define __exit<br> > +#define=C2=A0 =C2=A0 =C2=A0 __rcu<br> > +#define=C2=A0 =C2=A0 =C2=A0 __percpu<br> > +#define=C2=A0 =C2=A0 =C2=A0 __weak __weak_symbol<br> > +#define=C2=A0 =C2=A0 =C2=A0 __malloc<br> > +#define=C2=A0 =C2=A0 =C2=A0 ___stringify(...)=C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0#__VA_ARGS__<br> > +#define=C2=A0 =C2=A0 =C2=A0 __stringify(...)=C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 ___stringify(__VA_ARGS__)<br> > +#define=C2=A0 =C2=A0 =C2=A0 __attribute_const__=C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0__attribute__((__const__))<br> > +#undef __always_inline<br> > +#define=C2=A0 =C2=A0 =C2=A0 __always_inline=C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0inline<br> > +#define=C2=A0 =C2=A0 =C2=A0 noinline=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 __noinline<br> > +#define=C2=A0 =C2=A0 =C2=A0 ____cacheline_aligned=C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0__aligned(CACHE_LINE_SIZE)<br> > +<br> > +#define=C2=A0 =C2=A0 =C2=A0 likely(x)=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0__builtin_expect(!!(x),= 1)<br> > +#define=C2=A0 =C2=A0 =C2=A0 unlikely(x)=C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0__builtin_expect(!!(x), 0)<= br> > +#define typeof(x)=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 __typeof(x)<br> > +<br> > +#define=C2=A0 =C2=A0 =C2=A0 uninitialized_var(x)=C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 x =3D x<br> > +#define=C2=A0 =C2=A0 =C2=A0 __maybe_unused=C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 __unused<br> > +#define=C2=A0 =C2=A0 =C2=A0 __always_unused=C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0__unused<br> > +#define=C2=A0 =C2=A0 =C2=A0 __must_check=C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 __result_use_check<br> > +<br> > +#define=C2=A0 =C2=A0 =C2=A0 __printf(a,b)=C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0__printflike(a,b)<br> > +<br> > +#define=C2=A0 =C2=A0 =C2=A0 barrier()=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0__asm__ __volatile__(&q= uot;": : :"memory")<br> > +<br> > +#if defined(LINUXKPI_VERSION) && LINUXKPI_VERSION >=3D 500= 00<br> > +/* Moved from drm_os_freebsd.h */<br> > +#define=C2=A0 =C2=A0 =C2=A0 lower_32_bits(n)=C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 ((u32)(n))<br> > +#define=C2=A0 =C2=A0 =C2=A0 upper_32_bits(n)=C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 ((u32)(((n) >> 16) >> 16))<br> > +#endif<br> > +<br> > +#define=C2=A0 =C2=A0 =C2=A0 ___PASTE(a,b) a##b<br> > +#define=C2=A0 =C2=A0 =C2=A0 __PASTE(a,b) ___PASTE(a,b)<br> > +<br> > +#define=C2=A0 =C2=A0 =C2=A0 ACCESS_ONCE(x)=C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 (*(volatile __typeof(x) *)&(x))<br> > +<br> > +#define=C2=A0 =C2=A0 =C2=A0 WRITE_ONCE(x,v) do {=C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 \<br> > +=C2=A0 =C2=A0 =C2=A0barrier();=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 \<br> > +=C2=A0 =C2=A0 =C2=A0ACCESS_ONCE(x) =3D (v);=C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0\<br> > +=C2=A0 =C2=A0 =C2=A0barrier();=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 \<br> > +} while (0)<br> > +<br> > +#define=C2=A0 =C2=A0 =C2=A0 READ_ONCE(x) ({=C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0\<br> > +=C2=A0 =C2=A0 =C2=A0__typeof(x) __var =3D ({=C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 \<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0barrier();=C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 \<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0ACCESS_ONCE(x);=C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0\<br> > +=C2=A0 =C2=A0 =C2=A0});=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0\<br> > +=C2=A0 =C2=A0 =C2=A0barrier();=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 \<br> > +=C2=A0 =C2=A0 =C2=A0__var;=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 \<br> > +})<br> > +<br> > +#define=C2=A0 =C2=A0 =C2=A0 lockless_dereference(p) READ_ONCE(p)<br> > +<br> > +#define=C2=A0 =C2=A0 =C2=A0 _AT(T,X)=C2=A0 =C2=A0 =C2=A0 =C2=A0 ((T)(= X))<br> > +<br> > +#define=C2=A0 =C2=A0 =C2=A0 __same_type(a, b)=C2=A0 =C2=A0 =C2=A0 =C2= =A0__builtin_types_compatible_p(typeof(a), typeof(b))<br> > +#define=C2=A0 =C2=A0 =C2=A0 __must_be_array(a)=C2=A0 =C2=A0 =C2=A0 __= same_type(a, &(a)[0])<br> > +<br> > +#endif=C2=A0 =C2=A0 =C2=A0 =C2=A0/* _LINUX_COMPILER_H_ */<br> > diff --git a/sys/contrib/dev/iwlwifi/fw/api/soc.h b/sys/contrib/dev/iw= lwifi/fw/api/soc.h<br> > new file mode 100644<br> > index 000000000000..c5df1171462b<br> > --- /dev/null<br> > +++ b/sys/contrib/dev/iwlwifi/fw/api/soc.h<br> > @@ -0,0 +1,35 @@<br> > +/* SPDX-License-Identifier: GPL-2.0 OR BSD-3-Clause */<br> > +/*<br> > + * Copyright (C) 2012-2014, 2019-2020 Intel Corporation<br> > + * Copyright (C) 2013-2015 Intel Mobile Communications GmbH<br> > + * Copyright (C) 2016-2017 Intel Deutschland GmbH<br> > + */<br> > +#ifndef __iwl_fw_api_soc_h__<br> > +#define __iwl_fw_api_soc_h__<br> > +<br> > +#define SOC_CONFIG_CMD_FLAGS_DISCRETE=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 BIT(0)<br> > +#define SOC_CONFIG_CMD_FLAGS_LOW_LATENCY=C2=A0 =C2=A0 =C2=A0BIT(1)<br= > > +<br> > +#define SOC_FLAGS_LTR_APPLY_DELAY_MASK=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A00xc<br> > +#define SOC_FLAGS_LTR_APPLY_DELAY_NONE=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A00<br> > +#define SOC_FLAGS_LTR_APPLY_DELAY_200=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 1<br> > +#define SOC_FLAGS_LTR_APPLY_DELAY_2500=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A02<br> > +#define SOC_FLAGS_LTR_APPLY_DELAY_1820=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A03<br> > +<br> > +/**<br> > + * struct iwl_soc_configuration_cmd - Set device stabilization latenc= y<br> > + *<br> > + * @flags: soc settings flags.=C2=A0 In VER_1, we can only set the DI= SCRETE<br> > + *=C2=A0 =C2=A0flag, because the FW treats the whole value as an inte= ger. In<br> > + *=C2=A0 =C2=A0VER_2, we can set the bits independently.<br> > + * @latency: time for SOC to ensure stable power & XTAL<br> > + */<br> > +struct iwl_soc_configuration_cmd {<br> > +=C2=A0 =C2=A0 =C2=A0__le32 flags;<br> > +=C2=A0 =C2=A0 =C2=A0__le32 latency;<br> > +} __packed; /*<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 * SOC_CONFIGURATION_CMD_S_VER_1 (s= ee description above)<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 * SOC_CONFIGURATION_CMD_S_VER_2<br= > > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 */<br> > +<br> > +#endif /* __iwl_fw_api_soc_h__ */<br> > diff --git a/sys/contrib/zlib/contrib/asm686/README.686 b/sys/contrib/= zlib/contrib/asm686/README.686<br> > new file mode 100644<br> > index 000000000000..a0bf3bea4aff<br> > --- /dev/null<br> > +++ b/sys/contrib/zlib/contrib/asm686/README.686<br> > @@ -0,0 +1,51 @@<br> > +This is a patched version of zlib, modified to use<br> > +Pentium-Pro-optimized assembly code in the deflation algorithm. The<b= r> > +files changed/added by this patch are:<br> > +<br> > +README.686<br> > +match.S<br> > +<br> > +The speedup that this patch provides varies, depending on whether the= <br> > +compiler used to build the original version of zlib falls afoul of th= e<br> > +PPro's speed traps. My own tests show a speedup of around 10-20% = at<br> > +the default compression level, and 20-30% using -9, against a version= <br> > +compiled using gcc 2.7.2.3. Your mileage may vary.<br> > +<br> > +Note that this code has been tailored for the PPro/PII in particular,= <br> > +and will not perform particuarly well on a Pentium.<br> > +<br> > +If you are using an assembler other than GNU as, you will have to<br> > +translate match.S to use your assembler's syntax. (Have fun.)<br> > +<br> > +Brian Raiter<br> > +<a href=3D"mailto:breadbox@muppetlabs.com" target=3D"_blank" rel=3D"n= oreferrer">breadbox@muppetlabs.com</a><br> > +April, 1998<br> > +<br> > +<br> > +Added for zlib 1.1.3:<br> > +<br> > +The patches come from<br> > +<a href=3D"http://www.muppetlabs.com/~breadbox/software/assembly.html= " rel=3D"noreferrer noreferrer" target=3D"_blank">http://www.muppetlabs.com= /~breadbox/software/assembly.html</a><br> > +<br> > +To compile zlib with this asm file, copy match.S to the zlib director= y<br> > +then do:<br> > +<br> > +CFLAGS=3D"-O3 -DASMV" ./configure<br> > +make OBJA=3Dmatch.o<br> > +<br> > +<br> > +Update:<br> > +<br> > +I've been ignoring these assembly routines for years, believing t= hat<br> > +gcc's generated code had caught up with it sometime around gcc 2.= 95<br> > +and the major rearchitecting of the Pentium 4. However, I recently<br= > > +learned that, despite what I believed, this code still has some life<= br> > +in it. On the Pentium 4 and AMD64 chips, it continues to run about 8%= <br> > +faster than the code produced by gcc 4.1.<br> > +<br> > +In acknowledgement of its continuing usefulness, I've altered the= <br> > +license to match that of the rest of zlib. Share and Enjoy!<br> > +<br> > +Brian Raiter<br> > +<a href=3D"mailto:breadbox@muppetlabs.com" target=3D"_blank" rel=3D"n= oreferrer">breadbox@muppetlabs.com</a><br> > +April, 2007<br> > diff --git a/sys/contrib/zlib/contrib/asm686/match.S b/sys/contrib/zli= b/contrib/asm686/match.S<br> > new file mode 100644<br> > index 000000000000..fa421092785d<br> > --- /dev/null<br> > +++ b/sys/contrib/zlib/contrib/asm686/match.S<br> > @@ -0,0 +1,357 @@<br> > +/* match.S -- x86 assembly version of the zlib longest_match() functi= on.<br> > + * Optimized for the Intel 686 chips (PPro and later).<br> > + *<br> > + * Copyright (C) 1998, 2007 Brian Raiter <<a href=3D"mailto:breadb= ox@muppetlabs.com" target=3D"_blank" rel=3D"noreferrer">breadbox@muppetlabs= .com</a>><br> > + *<br> > + * This software is provided 'as-is', without any express or = implied<br> > + * warranty.=C2=A0 In no event will the author be held liable for any= damages<br> > + * arising from the use of this software.<br> > + *<br> > + * Permission is granted to anyone to use this software for any purpo= se,<br> > + * including commercial applications, and to alter it and redistribut= e it<br> > + * freely, subject to the following restrictions:<br> > + *<br> > + * 1. The origin of this software must not be misrepresented; you mus= t not<br> > + *=C2=A0 =C2=A0 claim that you wrote the original software. If you us= e this software<br> > + *=C2=A0 =C2=A0 in a product, an acknowledgment in the product docume= ntation would be<br> > + *=C2=A0 =C2=A0 appreciated but is not required.<br> > + * 2. Altered source versions must be plainly marked as such, and mus= t not be<br> > + *=C2=A0 =C2=A0 misrepresented as being the original software.<br> > + * 3. This notice may not be removed or altered from any source distr= ibution.<br> > + */<br> > +<br> > +#ifndef NO_UNDERLINE<br> > +#define=C2=A0 =C2=A0 =C2=A0 match_init=C2=A0 =C2=A0 =C2=A0 _match_ini= t<br> > +#define=C2=A0 =C2=A0 =C2=A0 longest_match=C2=A0 =C2=A0_longest_match<= br> > +#endif<br> > +<br> > +#define=C2=A0 =C2=A0 =C2=A0 MAX_MATCH=C2=A0 =C2=A0 =C2=A0 =C2=A0(258)= <br> > +#define=C2=A0 =C2=A0 =C2=A0 MIN_MATCH=C2=A0 =C2=A0 =C2=A0 =C2=A0(3)<b= r> > +#define=C2=A0 =C2=A0 =C2=A0 MIN_LOOKAHEAD=C2=A0 =C2=A0(MAX_MATCH + MI= N_MATCH + 1)<br> > +#define=C2=A0 =C2=A0 =C2=A0 MAX_MATCH_8=C2=A0 =C2=A0 =C2=A0((MAX_MATC= H + 7) & ~7)<br> > +<br> > +/* stack frame offsets */<br> > +<br> > +#define=C2=A0 =C2=A0 =C2=A0 chainlenwmask=C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A00=C2=A0 =C2=A0 =C2=A0 =C2=A0/* high word: current chain len */= <br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0/* low word:= s->wmask=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0*/<br> > +#define=C2=A0 =C2=A0 =C2=A0 window=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 4=C2=A0 =C2=A0 =C2=A0 =C2=A0/* local copy of s-= >window=C2=A0 =C2=A0 =C2=A0 */<br> > +#define=C2=A0 =C2=A0 =C2=A0 windowbestlen=C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A08=C2=A0 =C2=A0 =C2=A0 =C2=A0/* s->window + bestlen=C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 */<br> > +#define=C2=A0 =C2=A0 =C2=A0 scanstart=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A016=C2=A0 =C2=A0 =C2=A0 /* first two bytes of string= =C2=A0 =C2=A0 */<br> > +#define=C2=A0 =C2=A0 =C2=A0 scanend=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A012=C2=A0 =C2=A0 =C2=A0 /* last two bytes of str= ing=C2=A0 =C2=A0 =C2=A0*/<br> > +#define=C2=A0 =C2=A0 =C2=A0 scanalign=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A020=C2=A0 =C2=A0 =C2=A0 /* dword-misalignment of str= ing */<br> > +#define=C2=A0 =C2=A0 =C2=A0 nicematch=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A024=C2=A0 =C2=A0 =C2=A0 /* a good enough match size= =C2=A0 =C2=A0 =C2=A0*/<br> > +#define=C2=A0 =C2=A0 =C2=A0 bestlen=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A028=C2=A0 =C2=A0 =C2=A0 /* size of best match so= far=C2=A0 =C2=A0 */<br> > +#define=C2=A0 =C2=A0 =C2=A0 scan=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 32=C2=A0 =C2=A0 =C2=A0 /* ptr to string = wanting match=C2=A0 */<br> > +<br> > +#define=C2=A0 =C2=A0 =C2=A0 LocalVarsSize=C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0(36)<br> > +/*=C2=A0 =C2=A0saved ebx=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A036 */<br> > +/*=C2=A0 =C2=A0saved edi=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A040 */<br> > +/*=C2=A0 =C2=A0saved esi=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A044 */<br> > +/*=C2=A0 =C2=A0saved ebp=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A048 */<br> > +/*=C2=A0 =C2=A0return address=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 52 */= <br> > +#define=C2=A0 =C2=A0 =C2=A0 deflatestate=C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 56=C2=A0 =C2=A0 =C2=A0 /* the function arguments=C2=A0 =C2=A0= =C2=A0 =C2=A0*/<br> > +#define=C2=A0 =C2=A0 =C2=A0 curmatch=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 60<br> > +<br> > +/* All the +zlib1222add offsets are due to the addition of fields<br> > + *=C2=A0 in zlib in the deflate_state structure since the asm code wa= s first written<br> > + * (if you compile with zlib 1.0.4 or older, use "zlib1222add eq= u (-4)").<br> > + * (if you compile with zlib between 1.0.5 and 1.2.2.1, use "zli= b1222add equ 0").<br> > + * if you compile with zlib 1.2.2.2 or later , use "zlib1222add = equ 8").<br> > + */<br> > +<br> > +#define zlib1222add=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 (8)<br> > +<br> > +#define=C2=A0 =C2=A0 =C2=A0 dsWSize=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0(36+zlib1222add)<br> > +#define=C2=A0 =C2=A0 =C2=A0 dsWMask=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0(44+zlib1222add)<br> > +#define=C2=A0 =C2=A0 =C2=A0 dsWindow=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 (48+zlib1222add)<br> > +#define=C2=A0 =C2=A0 =C2=A0 dsPrev=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 (56+zlib1222add)<br> > +#define=C2=A0 =C2=A0 =C2=A0 dsMatchLen=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 (88+zlib1222add)<br> > +#define=C2=A0 =C2=A0 =C2=A0 dsPrevMatch=C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0(92+zlib1222add)<br> > +#define=C2=A0 =C2=A0 =C2=A0 dsStrStart=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 (100+zlib1222add)<br> > +#define=C2=A0 =C2=A0 =C2=A0 dsMatchStart=C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 (104+zlib1222add)<br> > +#define=C2=A0 =C2=A0 =C2=A0 dsLookahead=C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0(108+zlib1222add)<br> > +#define=C2=A0 =C2=A0 =C2=A0 dsPrevLen=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0(112+zlib1222add)<br> > +#define=C2=A0 =C2=A0 =C2=A0 dsMaxChainLen=C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0(116+zlib1222add)<br> > +#define=C2=A0 =C2=A0 =C2=A0 dsGoodMatch=C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0(132+zlib1222add)<br> > +#define=C2=A0 =C2=A0 =C2=A0 dsNiceMatch=C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0(136+zlib1222add)<br> > +<br> > +<br> > +.file "match.S"<br> > +<br> > +.globl=C2=A0 =C2=A0 =C2=A0 =C2=A0match_init, longest_match<br> > +<br> > +.text<br> > +<br> > +/* uInt longest_match(deflate_state *deflatestate, IPos curmatch) */<= br> > +.cfi_sections=C2=A0 =C2=A0 =C2=A0 =C2=A0 .debug_frame<br> > +<br> > +longest_match:<br> > +<br> > +.cfi_startproc<br> > +/* Save registers that the compiler may be using, and adjust %esp to = */<br> > +/* make room for our stack frame.=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 */<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0pushl=C2=A0 =C2=A0%eb= p<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0.cfi_def_cfa_offset 8= <br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0.cfi_offset ebp, -8<b= r> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0pushl=C2=A0 =C2=A0%ed= i<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0.cfi_def_cfa_offset 1= 2<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0pushl=C2=A0 =C2=A0%es= i<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0.cfi_def_cfa_offset 1= 6<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0pushl=C2=A0 =C2=A0%eb= x<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0.cfi_def_cfa_offset 2= 0<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0subl=C2=A0 =C2=A0 $Lo= calVarsSize, %esp<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0.cfi_def_cfa_offset L= ocalVarsSize+20<br> > +<br> > +/* Retrieve the function arguments. %ecx will hold cur_match=C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0*/<br> > +/* throughout the entire function. %edx will hold the pointer to the = */<br> > +/* deflate_state structure during the function's setup (before=C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0*/<br> > +/* entering the main loop).=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 */<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 def= latestate(%esp), %edx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 cur= match(%esp), %ecx<br> > +<br> > +/* uInt wmask =3D s->w_mask;=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0*/<br> > +/* unsigned chain_length =3D s->max_chain_length;=C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 */<br> > +/* if (s->prev_length >=3D s->good_match) {=C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 */<br> > +/*=C2=A0 =C2=A0 =C2=A0chain_length >>=3D 2;=C2=A0 =C2=A0 =C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0*/<br> > +/* }=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0*/<br> > + <br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 dsP= revLen(%edx), %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 dsG= oodMatch(%edx), %ebx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0cmpl=C2=A0 =C2=A0 %eb= x, %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 dsW= Mask(%edx), %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 dsM= axChainLen(%edx), %ebx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0jl=C2=A0 =C2=A0 =C2= =A0 LastMatchGood<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0shrl=C2=A0 =C2=A0 $2,= %ebx<br> > +LastMatchGood:<br> > +<br> > +/* chainlen is decremented once beforehand so that the function can= =C2=A0 */<br> > +/* use the sign flag instead of the zero flag for the exit test.=C2= =A0 =C2=A0 =C2=A0*/<br> > +/* It is then shifted into the high word, to make room for the wmask = */<br> > +/* value, which it will always accompany.=C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 */<br= > > +<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0decl=C2=A0 =C2=A0 %eb= x<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0shll=C2=A0 =C2=A0 $16= , %ebx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0orl=C2=A0 =C2=A0 =C2= =A0%eax, %ebx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 %eb= x, chainlenwmask(%esp)<br> > +<br> > +/* if ((uInt)nice_match > s->lookahead) nice_match =3D s->lo= okahead;=C2=A0 =C2=A0*/<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 dsN= iceMatch(%edx), %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 dsL= ookahead(%edx), %ebx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0cmpl=C2=A0 =C2=A0 %ea= x, %ebx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0jl=C2=A0 =C2=A0 =C2= =A0 LookaheadLess<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 %ea= x, %ebx<br> > +LookaheadLess:=C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 %ebx, nice= match(%esp)<br> > +<br> > +/* register Bytef *scan =3D s->window + s->strstart;=C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0*/<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 dsW= indow(%edx), %esi<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 %es= i, window(%esp)<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 dsS= trStart(%edx), %ebp<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0lea=C2=A0 =C2=A0 =C2= =A0(%esi,%ebp), %edi<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 %ed= i, scan(%esp)<br> > +<br> > +/* Determine how many bytes the scan ptr is off from being=C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0*/<br> > +/* dword-aligned.=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 */<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 %ed= i, %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0negl=C2=A0 =C2=A0 %ea= x<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0andl=C2=A0 =C2=A0 $3,= %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 %ea= x, scanalign(%esp)<br> > +<br> > +/* IPos limit =3D s->strstart > (IPos)MAX_DIST(s) ?=C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 */<br> > +/*=C2=A0 =C2=A0 =C2=A0s->strstart - (IPos)MAX_DIST(s) : NIL;=C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 */<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 dsW= Size(%edx), %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0subl=C2=A0 =C2=A0 $MI= N_LOOKAHEAD, %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0subl=C2=A0 =C2=A0 %ea= x, %ebp<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0jg=C2=A0 =C2=A0 =C2= =A0 LimitPositive<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0xorl=C2=A0 =C2=A0 %eb= p, %ebp<br> > +LimitPositive:<br> > +<br> > +/* int best_len =3D s->prev_length;=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 */<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 dsP= revLen(%edx), %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 %ea= x, bestlen(%esp)<br> > +<br> > +/* Store the sum of s->window + best_len in %esi locally, and in %= esi.=C2=A0 =C2=A0 =C2=A0 =C2=A0*/<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0addl=C2=A0 =C2=A0 %ea= x, %esi<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 %es= i, windowbestlen(%esp)<br> > +<br> > +/* register ush scan_start =3D *(ushf*)scan;=C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0*/= <br> > +/* register ush scan_end=C2=A0 =C2=A0=3D *(ushf*)(scan+best_len-1);= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 */<br> > +/* Posf *prev =3D s->prev;=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0*/<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movzwl=C2=A0 (%edi), = %ebx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 %eb= x, scanstart(%esp)<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movzwl=C2=A0 -1(%edi,= %eax), %ebx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 %eb= x, scanend(%esp)<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 dsP= rev(%edx), %edi<br> > +<br> > +/* Jump into the main loop.=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 */<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 cha= inlenwmask(%esp), %edx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0jmp=C2=A0 =C2=A0 =C2= =A0LoopEntry<br> > +<br> > +.balign 16<br> > +<br> > +/* do {<br> > + *=C2=A0 =C2=A0 =C2=A0match =3D s->window + cur_match;<br> > + *=C2=A0 =C2=A0 =C2=A0if (*(ushf*)(match+best_len-1) !=3D scan_end ||= <br> > + *=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0*(ushf*)match !=3D scan_start) co= ntinue;<br> > + *=C2=A0 =C2=A0 =C2=A0[...]<br> > + * } while ((cur_match =3D prev[cur_match & wmask]) > limit<br= > > + *=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 && --chain_length !=3D 0= );<br> > + *<br> > + * Here is the inner loop of the function. The function will spend th= e<br> > + * majority of its time in this loop, and majority of that time will<= br> > + * be spent in the first ten instructions.<br> > + *<br> > + * Within this loop:<br> > + * %ebx =3D scanend<br> > + * %ecx =3D curmatch<br> > + * %edx =3D chainlenwmask - i.e., ((chainlen << 16) | wmask)<br= > > + * %esi =3D windowbestlen - i.e., (window + bestlen)<br> > + * %edi =3D prev<br> > + * %ebp =3D limit<br> > + */<br> > +LookupLoop:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0andl=C2=A0 =C2=A0 %ed= x, %ecx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movzwl=C2=A0 (%edi,%e= cx,2), %ecx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0cmpl=C2=A0 =C2=A0 %eb= p, %ecx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0jbe=C2=A0 =C2=A0 =C2= =A0LeaveNow<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0subl=C2=A0 =C2=A0 $0x= 00010000, %edx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0js=C2=A0 =C2=A0 =C2= =A0 LeaveNow<br> > +LoopEntry:=C2=A0 =C2=A0movzwl=C2=A0 -1(%esi,%ecx), %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0cmpl=C2=A0 =C2=A0 %eb= x, %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0jnz=C2=A0 =C2=A0 =C2= =A0LookupLoop<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 win= dow(%esp), %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movzwl=C2=A0 (%eax,%e= cx), %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0cmpl=C2=A0 =C2=A0 sca= nstart(%esp), %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0jnz=C2=A0 =C2=A0 =C2= =A0LookupLoop<br> > +<br> > +/* Store the current value of chainlen.=C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 */<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 %ed= x, chainlenwmask(%esp)<br> > +<br> > +/* Point %edi to the string under scrutiny, and %esi to the string we= =C2=A0 =C2=A0 =C2=A0 =C2=A0 */<br> > +/* are hoping to match it up with. In actuality, %esi and %edi are=C2= =A0 =C2=A0*/<br> > +/* both pointed (MAX_MATCH_8 - scanalign) bytes ahead, and %edx is=C2= =A0 =C2=A0*/<br> > +/* initialized to -(MAX_MATCH_8 - scanalign).=C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 */<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 win= dow(%esp), %esi<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 sca= n(%esp), %edi<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0addl=C2=A0 =C2=A0 %ec= x, %esi<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 sca= nalign(%esp), %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 $(-= MAX_MATCH_8), %edx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0lea=C2=A0 =C2=A0 =C2= =A0MAX_MATCH_8(%edi,%eax), %edi<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0lea=C2=A0 =C2=A0 =C2= =A0MAX_MATCH_8(%esi,%eax), %esi<br> > +<br> > +/* Test the strings for equality, 8 bytes at a time. At the end,<br> > + * adjust %edx so that it is offset to the exact byte that mismatched= .<br> > + *<br> > + * We already know at this point that the first three bytes of the<br= > > + * strings match each other, and they can be safely passed over befor= e<br> > + * starting the compare loop. So what this code does is skip over 0-3= <br> > + * bytes, as much as necessary in order to dword-align the %edi<br> > + * pointer. (%esi will still be misaligned three times out of four.)<= br> > + *<br> > + * It should be confessed that this loop usually does not represent<b= r> > + * much of the total running time. Replacing it with a more<br> > + * straightforward "rep cmpsb" would not drastically degrad= e<br> > + * performance.<br> > + */<br> > +LoopCmps:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 (%e= si,%edx), %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0xorl=C2=A0 =C2=A0 (%e= di,%edx), %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0jnz=C2=A0 =C2=A0 =C2= =A0LeaveLoopCmps<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 4(%= esi,%edx), %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0xorl=C2=A0 =C2=A0 4(%= edi,%edx), %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0jnz=C2=A0 =C2=A0 =C2= =A0LeaveLoopCmps4<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0addl=C2=A0 =C2=A0 $8,= %edx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0jnz=C2=A0 =C2=A0 =C2= =A0LoopCmps<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0jmp=C2=A0 =C2=A0 =C2= =A0LenMaximum<br> > +LeaveLoopCmps4:=C2=A0 =C2=A0 =C2=A0 addl=C2=A0 =C2=A0 $4, %edx<br> > +LeaveLoopCmps:=C2=A0 =C2=A0 =C2=A0 =C2=A0testl=C2=A0 =C2=A0$0x0000FFF= F, %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0jnz=C2=A0 =C2=A0 =C2= =A0LenLower<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0addl=C2=A0 =C2=A0 $2,= %edx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0shrl=C2=A0 =C2=A0 $16= , %eax<br> > +LenLower:=C2=A0 =C2=A0 subb=C2=A0 =C2=A0 $1, %al<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0adcl=C2=A0 =C2=A0 $0,= %edx<br> > +<br> > +/* Calculate the length of the match. If it is longer than MAX_MATCH,= =C2=A0 =C2=A0 =C2=A0 =C2=A0 */<br> > +/* then automatically accept it as the best possible match and leave.= =C2=A0 =C2=A0 =C2=A0 =C2=A0 */<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0lea=C2=A0 =C2=A0 =C2= =A0(%edi,%edx), %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 sca= n(%esp), %edi<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0subl=C2=A0 =C2=A0 %ed= i, %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0cmpl=C2=A0 =C2=A0 $MA= X_MATCH, %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0jge=C2=A0 =C2=A0 =C2= =A0LenMaximum<br> > +<br> > +/* If the length of the match is not longer than the best match we=C2= =A0 =C2=A0*/<br> > +/* have so far, then forget it and return to the lookup loop.=C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 */<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 def= latestate(%esp), %edx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 bes= tlen(%esp), %ebx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0cmpl=C2=A0 =C2=A0 %eb= x, %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0jg=C2=A0 =C2=A0 =C2= =A0 LongerMatch<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 win= dowbestlen(%esp), %esi<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 dsP= rev(%edx), %edi<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 sca= nend(%esp), %ebx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 cha= inlenwmask(%esp), %edx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0jmp=C2=A0 =C2=A0 =C2= =A0LookupLoop<br> > +<br> > +/*=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0s->match_start =3D cur_match;= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0*/<br> > +/*=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0best_len =3D len;=C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0*/<br> > +/*=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0if (len >=3D nice_match) break= ;=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0*/<br> > +/*=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0scan_end =3D *(ushf*)(scan+best_l= en-1);=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0*/<br> > +<br> > +LongerMatch: movl=C2=A0 =C2=A0 nicematch(%esp), %ebx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 %ea= x, bestlen(%esp)<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 %ec= x, dsMatchStart(%edx)<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0cmpl=C2=A0 =C2=A0 %eb= x, %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0jge=C2=A0 =C2=A0 =C2= =A0LeaveNow<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 win= dow(%esp), %esi<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0addl=C2=A0 =C2=A0 %ea= x, %esi<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 %es= i, windowbestlen(%esp)<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movzwl=C2=A0 -1(%edi,= %eax), %ebx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 dsP= rev(%edx), %edi<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 %eb= x, scanend(%esp)<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 cha= inlenwmask(%esp), %edx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0jmp=C2=A0 =C2=A0 =C2= =A0LookupLoop<br> > +<br> > +/* Accept the current string, with the maximum possible length.=C2=A0= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 */<br> > +<br> > +LenMaximum:=C2=A0 movl=C2=A0 =C2=A0 deflatestate(%esp), %edx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 $MA= X_MATCH, bestlen(%esp)<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 %ec= x, dsMatchStart(%edx)<br> > +<br> > +/* if ((uInt)best_len <=3D s->lookahead) return (uInt)best_len;= =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 */<br> > +/* return s->lookahead;=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 = =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 *= /<br> > +<br> > +LeaveNow:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 def= latestate(%esp), %edx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 bes= tlen(%esp), %ebx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 dsL= ookahead(%edx), %eax<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0cmpl=C2=A0 =C2=A0 %ea= x, %ebx<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0jg=C2=A0 =C2=A0 =C2= =A0 LookaheadRet<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0movl=C2=A0 =C2=A0 %eb= x, %eax<br> > +LookaheadRet:<br> > +<br> > +/* Restore the stack and return from whence we came.=C2=A0 =C2=A0 =C2= =A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0*/<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0addl=C2=A0 =C2=A0 $Lo= calVarsSize, %esp<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0.cfi_def_cfa_offset 2= 0<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0popl=C2=A0 =C2=A0 %eb= x<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0.cfi_def_cfa_offset 1= 6<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0popl=C2=A0 =C2=A0 %es= i<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0.cfi_def_cfa_offset 1= 2<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0popl=C2=A0 =C2=A0 %ed= i<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0.cfi_def_cfa_offset 8= <br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0popl=C2=A0 =C2=A0 %eb= p<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0.cfi_def_cfa_offset 4= <br> > +.cfi_endproc<br> > +match_init:=C2=A0 ret<br> > diff --git a/sys/dev/ice/ice_sriov.c b/sys/dev/ice/ice_sriov.c<br> > new file mode 100644<br> > index 000000000000..c0521e667fa2<br> > --- /dev/null<br> > +++ b/sys/dev/ice/ice_sriov.c<br> > @@ -0,0 +1,595 @@<br> > +/* SPDX-License-Identifier: BSD-3-Clause */<br> > +/*=C2=A0 Copyright (c) 2021, Intel Corporation<br> > + *=C2=A0 All rights reserved.<br> > + *<br> > + *=C2=A0 Redistribution and use in source and binary forms, with or w= ithout<br> > + *=C2=A0 modification, are permitted provided that the following cond= itions are met:<br> > + *<br> > + *=C2=A0 =C2=A01. Redistributions of source code must retain the abov= e copyright notice,<br> > + *=C2=A0 =C2=A0 =C2=A0 this list of conditions and the following disc= laimer.<br> > + *<br> > + *=C2=A0 =C2=A02. Redistributions in binary form must reproduce the a= bove copyright<br> > + *=C2=A0 =C2=A0 =C2=A0 notice, this list of conditions and the follow= ing disclaimer in the<br> > + *=C2=A0 =C2=A0 =C2=A0 documentation and/or other materials provided = with the distribution.<br> > + *<br> > + *=C2=A0 =C2=A03. Neither the name of the Intel Corporation nor the n= ames of its<br> > + *=C2=A0 =C2=A0 =C2=A0 contributors may be used to endorse or promote= products derived from<br> > + *=C2=A0 =C2=A0 =C2=A0 this software without specific prior written p= ermission.<br> > + *<br> > + *=C2=A0 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTR= IBUTORS "AS IS"<br> > + *=C2=A0 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LI= MITED TO, THE<br> > + *=C2=A0 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PART= ICULAR PURPOSE<br> > + *=C2=A0 ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CON= TRIBUTORS BE<br> > + *=C2=A0 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPL= ARY, OR<br> > + *=C2=A0 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUR= EMENT OF<br> > + *=C2=A0 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;= OR BUSINESS<br> > + *=C2=A0 INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,= WHETHER IN<br> > + *=C2=A0 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR= OTHERWISE)<br> > + *=C2=A0 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF = ADVISED OF THE<br> > + *=C2=A0 POSSIBILITY OF SUCH DAMAGE.<br> > + */<br> > +<br> > +#include "ice_common.h"<br> > +#include "ice_sriov.h"<br> > +<br> > +/**<br> > + * ice_aq_send_msg_to_vf<br> > + * @hw: pointer to the hardware structure<br> > + * @vfid: VF ID to send msg<br> > + * @v_opcode: opcodes for VF-PF communication<br> > + * @v_retval: return error code<br> > + * @msg: pointer to the msg buffer<br> > + * @msglen: msg length<br> > + * @cd: pointer to command details<br> > + *<br> > + * Send message to VF driver (0x0802) using mailbox<br> > + * queue and asynchronously sending message via<br> > + * ice_sq_send_cmd() function<br> > + */<br> > +enum ice_status<br> > +ice_aq_send_msg_to_vf(struct ice_hw *hw, u16 vfid, u32 v_opcode, u32 = v_retval,<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= u8 *msg, u16 msglen, struct ice_sq_cd *cd)<br> > +{<br> > +=C2=A0 =C2=A0 =C2=A0struct ice_aqc_pf_vf_msg *cmd;<br> > +=C2=A0 =C2=A0 =C2=A0struct ice_aq_desc desc;<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0ice_fill_dflt_direct_cmd_desc(&desc, ice_mbx_= opc_send_msg_to_vf);<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0cmd =3D &desc.params.virt;<br> > +=C2=A0 =C2=A0 =C2=A0cmd->id =3D CPU_TO_LE32(vfid);<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0desc.cookie_high =3D CPU_TO_LE32(v_opcode);<br> > +=C2=A0 =C2=A0 =C2=A0desc.cookie_low =3D CPU_TO_LE32(v_retval);<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0if (msglen)<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0desc.flags |=3D CPU_T= O_LE16(ICE_AQ_FLAG_RD);<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0return ice_sq_send_cmd(hw, &hw->mailboxq, = &desc, msg, msglen, cd);<br> > +}<br> > +<br> > +/**<br> > + * ice_aq_send_msg_to_pf<br> > + * @hw: pointer to the hardware structure<br> > + * @v_opcode: opcodes for VF-PF communication<br> > + * @v_retval: return error code<br> > + * @msg: pointer to the msg buffer<br> > + * @msglen: msg length<br> > + * @cd: pointer to command details<br> > + *<br> > + * Send message to PF driver using mailbox queue. By default, this<br= > > + * message is sent asynchronously, i.e. ice_sq_send_cmd()<br> > + * does not wait for completion before returning.<br> > + */<br> > +enum ice_status<br> > +ice_aq_send_msg_to_pf(struct ice_hw *hw, enum virtchnl_ops v_opcode,<= br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= enum ice_status v_retval, u8 *msg, u16 msglen,<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= struct ice_sq_cd *cd)<br> > +{<br> > +=C2=A0 =C2=A0 =C2=A0struct ice_aq_desc desc;<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0ice_fill_dflt_direct_cmd_desc(&desc, ice_mbx_= opc_send_msg_to_pf);<br> > +=C2=A0 =C2=A0 =C2=A0desc.cookie_high =3D CPU_TO_LE32(v_opcode);<br> > +=C2=A0 =C2=A0 =C2=A0desc.cookie_low =3D CPU_TO_LE32(v_retval);<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0if (msglen)<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0desc.flags |=3D CPU_T= O_LE16(ICE_AQ_FLAG_RD);<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0return ice_sq_send_cmd(hw, &hw->mailboxq, = &desc, msg, msglen, cd);<br> > +}<br> > +<br> > +/**<br> > + * ice_conv_link_speed_to_virtchnl<br> > + * @adv_link_support: determines the format of the returned link spee= d<br> > + * @link_speed: variable containing the link_speed to be converted<br= > > + *<br> > + * Convert link speed supported by HW to link speed supported by virt= chnl.<br> > + * If adv_link_support is true, then return link speed in Mbps. Else = return<br> > + * link speed as a VIRTCHNL_LINK_SPEED_* casted to a u32. Note that t= he caller<br> > + * needs to cast back to an enum virtchnl_link_speed in the case wher= e<br> > + * adv_link_support is false, but when adv_link_support is true the c= aller can<br> > + * expect the speed in Mbps.<br> > + */<br> > +u32 ice_conv_link_speed_to_virtchnl(bool adv_link_support, u16 link_s= peed)<br> > +{<br> > +=C2=A0 =C2=A0 =C2=A0u32 speed;<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0if (adv_link_support)<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0switch (link_speed) {= <br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_10MB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0speed =3D ICE_LINK_SPEED_10MBPS;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0break;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_100MB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0speed =3D ICE_LINK_SPEED_100MBPS;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0break;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_1000MB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0speed =3D ICE_LINK_SPEED_1000MBPS;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0break;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_2500MB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0speed =3D ICE_LINK_SPEED_2500MBPS;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0break;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_5GB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0speed =3D ICE_LINK_SPEED_5000MBPS;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0break;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_10GB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0speed =3D ICE_LINK_SPEED_10000MBPS;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0break;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_20GB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0speed =3D ICE_LINK_SPEED_20000MBPS;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0break;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_25GB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0speed =3D ICE_LINK_SPEED_25000MBPS;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0break;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_40GB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0speed =3D ICE_LINK_SPEED_40000MBPS;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0break;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_50GB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0speed =3D ICE_LINK_SPEED_50000MBPS;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0break;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_100GB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0speed =3D ICE_LINK_SPEED_100000MBPS;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0break;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0default:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0speed =3D ICE_LINK_SPEED_UNKNOWN;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0break;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0}<br> > +=C2=A0 =C2=A0 =C2=A0else<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0/* Virtchnl speeds ar= e not defined for every speed supported in<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 * the hardware. To m= aintain compatibility with older AVF<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 * drivers, while rep= orting the speed the new speed values are<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 * resolved to the cl= osest known virtchnl speeds<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 */<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0switch (link_speed) {= <br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_10MB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_100MB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0speed =3D (u32)VIRTCHNL_LINK_SPEED_100MB;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0break;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_1000MB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_2500MB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_5GB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0speed =3D (u32)VIRTCHNL_LINK_SPEED_1GB;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0break;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_10GB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0speed =3D (u32)VIRTCHNL_LINK_SPEED_10GB;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0break;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_20GB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0speed =3D (u32)VIRTCHNL_LINK_SPEED_20GB;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0break;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_25GB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0speed =3D (u32)VIRTCHNL_LINK_SPEED_25GB;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0break;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_40GB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_50GB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0case ICE_AQ_LINK_SPEE= D_100GB:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0speed =3D (u32)VIRTCHNL_LINK_SPEED_40GB;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0break;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0default:<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0speed =3D (u32)VIRTCHNL_LINK_SPEED_UNKNOWN;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0= =C2=A0break;<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0}<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0return speed;<br> > +}<br> > +<br> > +/* The mailbox overflow detection algorithm helps to check if there<b= r> > + * is a possibility of a malicious VF transmitting too many MBX messa= ges to the<br> > + * PF.<br> > + * 1. The mailbox snapshot structure, ice_mbx_snapshot, is initialize= d during<br> > + * driver initialization in ice_init_hw() using ice_mbx_init_snapshot= ().<br> > + * The struct ice_mbx_snapshot helps to track and traverse a static w= indow of<br> > + * messages within the mailbox queue while looking for a malicious VF= .<br> > + *<br> > + * 2. When the caller starts processing its mailbox queue in response= to an<br> > + * interrupt, the structure ice_mbx_snapshot is expected to be cleare= d before<br> > + * the algorithm can be run for the first time for that interrupt. Th= is can be<br> > + * done via ice_mbx_reset_snapshot().<br> > + *<br> > + * 3. For every message read by the caller from the MBX Queue, the ca= ller must<br> > + * call the detection algorithm's entry function ice_mbx_vf_state= _handler().<br> > + * Before every call to ice_mbx_vf_state_handler() the struct ice_mbx= _data is<br> > + * filled as it is required to be passed to the algorithm.<br> > + *<br> > + * 4. Every time a message is read from the MBX queue, a VFId is rece= ived which<br> > + * is passed to the state handler. The boolean output is_malvf of the= state<br> > + * handler ice_mbx_vf_state_handler() serves as an indicator to the c= aller<br> > + * whether this VF is malicious or not.<br> > + *<br> > + * 5. When a VF is identified to be malicious, the caller can send a = message<br> > + * to the system administrator. The caller can invoke ice_mbx_report_= malvf()<br> > + * to help determine if a malicious VF is to be reported or not. This= function<br> > + * requires the caller to maintain a global bitmap to track all malic= ious VFs<br> > + * and pass that to ice_mbx_report_malvf() along with the VFID which = was identified<br> > + * to be malicious by ice_mbx_vf_state_handler().<br> > + *<br> > + * 6. The global bitmap maintained by PF can be cleared completely if= PF is in<br> > + * reset or the bit corresponding to a VF can be cleared if that VF i= s in reset.<br> > + * When a VF is shut down and brought back up, we assume that the new= VF<br> > + * brought up is not malicious and hence report it if found malicious= .<br> > + *<br> > + * 7. The function ice_mbx_reset_snapshot() is called to reset the in= formation<br> > + * in ice_mbx_snapshot for every new mailbox interrupt handled.<br> > + *<br> > + * 8. The memory allocated for variables in ice_mbx_snapshot is de-al= located<br> > + * when driver is unloaded.<br> > + */<br> > +#define ICE_RQ_DATA_MASK(rq_data) ((rq_data) & PF_MBX_ARQH_ARQH_M= )<br> > +/* Using the highest value for an unsigned 16-bit value 0xFFFF to ind= icate that<br> > + * the max messages check must be ignored in the algorithm<br> > + */<br> > +#define ICE_IGNORE_MAX_MSG_CNT=C2=A0 =C2=A0 =C2=A0 =C2=A00xFFFF<br> > +<br> > +/**<br> > + * ice_mbx_traverse - Pass through mailbox snapshot<br> > + * @hw: pointer to the HW struct<br> > + * @new_state: new algorithm state<br> > + *<br> > + * Traversing the mailbox static snapshot without checking<br> > + * for malicious VFs.<br> > + */<br> > +static void<br> > +ice_mbx_traverse(struct ice_hw *hw,<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 enum ice_mbx_snapsho= t_state *new_state)<br> > +{<br> > +=C2=A0 =C2=A0 =C2=A0struct ice_mbx_snap_buffer_data *snap_buf;<br> > +=C2=A0 =C2=A0 =C2=A0u32 num_iterations;<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0snap_buf =3D &hw->mbx_snapshot.mbx_buf;<br= > > +<br> > +=C2=A0 =C2=A0 =C2=A0/* As mailbox buffer is circular, applying a mask= <br> > +=C2=A0 =C2=A0 =C2=A0 * on the incremented iteration count.<br> > +=C2=A0 =C2=A0 =C2=A0 */<br> > +=C2=A0 =C2=A0 =C2=A0num_iterations =3D ICE_RQ_DATA_MASK(++snap_buf-&g= t;num_iterations);<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0/* Checking either of the below conditions to exi= t snapshot traversal:<br> > +=C2=A0 =C2=A0 =C2=A0 * Condition-1: If the number of iterations in th= e mailbox is equal to<br> > +=C2=A0 =C2=A0 =C2=A0 * the mailbox head which would indicate that we = have reached the end<br> > +=C2=A0 =C2=A0 =C2=A0 * of the static snapshot.<br> > +=C2=A0 =C2=A0 =C2=A0 * Condition-2: If the maximum messages serviced = in the mailbox for a<br> > +=C2=A0 =C2=A0 =C2=A0 * given interrupt is the highest possible value = then there is no need<br> > +=C2=A0 =C2=A0 =C2=A0 * to check if the number of messages processed i= s equal to it. If not<br> > +=C2=A0 =C2=A0 =C2=A0 * check if the number of messages processed is g= reater than or equal<br> > +=C2=A0 =C2=A0 =C2=A0 * to the maximum number of mailbox entries servi= ced in current work item.<br> > +=C2=A0 =C2=A0 =C2=A0 */<br> > +=C2=A0 =C2=A0 =C2=A0if (num_iterations =3D=3D snap_buf->head ||<br= > > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0(snap_buf->max_num_msgs_mbx <= ICE_IGNORE_MAX_MSG_CNT &&<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 ++snap_buf->num_msg_proc >= =3D snap_buf->max_num_msgs_mbx))<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0*new_state =3D ICE_MA= L_VF_DETECT_STATE_NEW_SNAPSHOT;<br> > +}<br> > +<br> > +/**<br> > + * ice_mbx_detect_malvf - Detect malicious VF in snapshot<br> > + * @hw: pointer to the HW struct<br> > + * @vf_id: relative virtual function ID<br> > + * @new_state: new algorithm state<br> > + * @is_malvf: boolean output to indicate if VF is malicious<br> > + *<br> > + * This function tracks the number of asynchronous messages<br> > + * sent per VF and marks the VF as malicious if it exceeds<br> > + * the permissible number of messages to send.<br> > + */<br> > +static enum ice_status<br> > +ice_mbx_detect_malvf(struct ice_hw *hw, u16 vf_id,<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 enum i= ce_mbx_snapshot_state *new_state,<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 bool *= is_malvf)<br> > +{<br> > +=C2=A0 =C2=A0 =C2=A0struct ice_mbx_snapshot *snap =3D &hw->mbx= _snapshot;<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0if (vf_id >=3D snap->mbx_vf.vfcntr_len)<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0return ICE_ERR_OUT_OF= _RANGE;<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0/* increment the message count in the VF array */= <br> > +=C2=A0 =C2=A0 =C2=A0snap->mbx_vf.vf_cntr[vf_id]++;<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0if (snap->mbx_vf.vf_cntr[vf_id] >=3D ICE_AS= YNC_VF_MSG_THRESHOLD)<br> > +=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0*is_malvf =3D true;<b= r> > +<br> > +=C2=A0 =C2=A0 =C2=A0/* continue to iterate through the mailbox snapsh= ot */<br> > +=C2=A0 =C2=A0 =C2=A0ice_mbx_traverse(hw, new_state);<br> > +<br> > +=C2=A0 =C2=A0 =C2=A0return ICE_SUCCESS;<br> > +}<br> > +<br> > +/**<br> > + * ice_mbx_reset_snapshot - Reset mailbox snapshot structure<br> > + * @snap: pointer to mailbox snapshot structure in the ice_hw struct<= br> > + *<br> > + * Reset the mailbox snapshot structure and clear VF counter array.<b= r> > + */<br> > +static void ice_mbx_reset_snapshot(struct ice_mbx_snapshot *snap)<br> > *** 15611 LINES SKIPPED ***<br> <br> <br> <br> </blockquote></div></div></div> --000000000000014480061fc48d49--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CANCZdfpk7dSiuYaU42cFXrhqMsonbPrQrjzTDeWL=vvCnoHMxA>