Init ipxe source code cloned fron ipxe.org at Oct 29th 2020

.travis.yml

@@ -0,0 +1,57 @@
+dist: trusty
+
+sudo: false
+
+git:
+  depth: false
+
+language: c
+
+cache: ccache
+
+compiler:
+  - gcc
+
+addons:
+  apt:
+    packages:
+      - binutils-dev
+      - liblzma-dev
+      - syslinux
+      - genisoimage
+  coverity_scan:
+    project:
+      name: "ipxe/ipxe"
+      version: $TRAVIS_COMMIT
+    build_command_prepend: "make -C src bin/deps"
+    build_command: "make -C src bin/blib.a"
+    branch_pattern: coverity_scan
+
+env:
+  global:
+    - MAKEFLAGS="-j 4"
+
+script:
+  - make -C src bin/blib.a
+  - make -C src bin/ipxe.pxe
+  - make -C src bin/ipxe.usb
+  - make -C src bin/ipxe.iso
+  - make -C src bin/8086100e.mrom
+  - make -C src bin-x86_64-pcbios/blib.a
+  - make -C src bin-x86_64-pcbios/ipxe.pxe
+  - make -C src bin-x86_64-pcbios/ipxe.usb
+  - make -C src bin-x86_64-pcbios/ipxe.iso
+  - make -C src bin-x86_64-pcbios/8086100e.mrom
+  - make -C src bin-x86_64-efi/blib.a
+  - make -C src bin-x86_64-efi/ipxe.efi
+  - make -C src bin-x86_64-efi/intel.efidrv
+  - make -C src bin-x86_64-efi/intel.efirom
+  - make -C src bin-i386-efi/blib.a
+  - make -C src bin-i386-efi/ipxe.efi
+  - make -C src bin-i386-efi/intel.efidrv
+  - make -C src bin-i386-efi/intel.efirom
+  - make -C src bin-x86_64-linux/blib.a
+  - make -C src bin-x86_64-linux/tap.linux
+  - make -C src bin-x86_64-linux/af_packet.linux
+  - make -C src bin-x86_64-linux/tests.linux
+  - ./src/bin-x86_64-linux/tests.linux

COPYING.GPLv2

@@ -0,0 +1,339 @@
+                    GNU GENERAL PUBLIC LICENSE
+                       Version 2, June 1991
+
+ Copyright (C) 1989, 1991 Free Software Foundation, Inc.,
+ 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ Everyone is permitted to copy and distribute verbatim copies
+ of this license document, but changing it is not allowed.
+
+                            Preamble
+
+  The licenses for most software are designed to take away your
+freedom to share and change it.  By contrast, the GNU General Public
+License is intended to guarantee your freedom to share and change free
+software--to make sure the software is free for all its users.  This
+General Public License applies to most of the Free Software
+Foundation's software and to any other program whose authors commit to
+using it.  (Some other Free Software Foundation software is covered by
+the GNU Lesser General Public License instead.)  You can apply it to
+your programs, too.
+
+  When we speak of free software, we are referring to freedom, not
+price.  Our General Public Licenses are designed to make sure that you
+have the freedom to distribute copies of free software (and charge for
+this service if you wish), that you receive source code or can get it
+if you want it, that you can change the software or use pieces of it
+in new free programs; and that you know you can do these things.
+
+  To protect your rights, we need to make restrictions that forbid
+anyone to deny you these rights or to ask you to surrender the rights.
+These restrictions translate to certain responsibilities for you if you
+distribute copies of the software, or if you modify it.
+
+  For example, if you distribute copies of such a program, whether
+gratis or for a fee, you must give the recipients all the rights that
+you have.  You must make sure that they, too, receive or can get the
+source code.  And you must show them these terms so they know their
+rights.
+
+  We protect your rights with two steps: (1) copyright the software, and
+(2) offer you this license which gives you legal permission to copy,
+distribute and/or modify the software.
+
+  Also, for each author's protection and ours, we want to make certain
+that everyone understands that there is no warranty for this free
+software.  If the software is modified by someone else and passed on, we
+want its recipients to know that what they have is not the original, so
+that any problems introduced by others will not reflect on the original
+authors' reputations.
+
+  Finally, any free program is threatened constantly by software
+patents.  We wish to avoid the danger that redistributors of a free
+program will individually obtain patent licenses, in effect making the
+program proprietary.  To prevent this, we have made it clear that any
+patent must be licensed for everyone's free use or not licensed at all.
+
+  The precise terms and conditions for copying, distribution and
+modification follow.
+
+                    GNU GENERAL PUBLIC LICENSE
+   TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
+
+  0. This License applies to any program or other work which contains
+a notice placed by the copyright holder saying it may be distributed
+under the terms of this General Public License.  The "Program", below,
+refers to any such program or work, and a "work based on the Program"
+means either the Program or any derivative work under copyright law:
+that is to say, a work containing the Program or a portion of it,
+either verbatim or with modifications and/or translated into another
+language.  (Hereinafter, translation is included without limitation in
+the term "modification".)  Each licensee is addressed as "you".
+
+Activities other than copying, distribution and modification are not
+covered by this License; they are outside its scope.  The act of
+running the Program is not restricted, and the output from the Program
+is covered only if its contents constitute a work based on the
+Program (independent of having been made by running the Program).
+Whether that is true depends on what the Program does.
+
+  1. You may copy and distribute verbatim copies of the Program's
+source code as you receive it, in any medium, provided that you
+conspicuously and appropriately publish on each copy an appropriate
+copyright notice and disclaimer of warranty; keep intact all the
+notices that refer to this License and to the absence of any warranty;
+and give any other recipients of the Program a copy of this License
+along with the Program.
+
+You may charge a fee for the physical act of transferring a copy, and
+you may at your option offer warranty protection in exchange for a fee.
+
+  2. You may modify your copy or copies of the Program or any portion
+of it, thus forming a work based on the Program, and copy and
+distribute such modifications or work under the terms of Section 1
+above, provided that you also meet all of these conditions:
+
+    a) You must cause the modified files to carry prominent notices
+    stating that you changed the files and the date of any change.
+
+    b) You must cause any work that you distribute or publish, that in
+    whole or in part contains or is derived from the Program or any
+    part thereof, to be licensed as a whole at no charge to all third
+    parties under the terms of this License.
+
+    c) If the modified program normally reads commands interactively
+    when run, you must cause it, when started running for such
+    interactive use in the most ordinary way, to print or display an
+    announcement including an appropriate copyright notice and a
+    notice that there is no warranty (or else, saying that you provide
+    a warranty) and that users may redistribute the program under
+    these conditions, and telling the user how to view a copy of this
+    License.  (Exception: if the Program itself is interactive but
+    does not normally print such an announcement, your work based on
+    the Program is not required to print an announcement.)
+
+These requirements apply to the modified work as a whole.  If
+identifiable sections of that work are not derived from the Program,
+and can be reasonably considered independent and separate works in
+themselves, then this License, and its terms, do not apply to those
+sections when you distribute them as separate works.  But when you
+distribute the same sections as part of a whole which is a work based
+on the Program, the distribution of the whole must be on the terms of
+this License, whose permissions for other licensees extend to the
+entire whole, and thus to each and every part regardless of who wrote it.
+
+Thus, it is not the intent of this section to claim rights or contest
+your rights to work written entirely by you; rather, the intent is to
+exercise the right to control the distribution of derivative or
+collective works based on the Program.
+
+In addition, mere aggregation of another work not based on the Program
+with the Program (or with a work based on the Program) on a volume of
+a storage or distribution medium does not bring the other work under
+the scope of this License.
+
+  3. You may copy and distribute the Program (or a work based on it,
+under Section 2) in object code or executable form under the terms of
+Sections 1 and 2 above provided that you also do one of the following:
+
+    a) Accompany it with the complete corresponding machine-readable
+    source code, which must be distributed under the terms of Sections
+    1 and 2 above on a medium customarily used for software interchange; or,
+
+    b) Accompany it with a written offer, valid for at least three
+    years, to give any third party, for a charge no more than your
+    cost of physically performing source distribution, a complete
+    machine-readable copy of the corresponding source code, to be
+    distributed under the terms of Sections 1 and 2 above on a medium
+    customarily used for software interchange; or,
+
+    c) Accompany it with the information you received as to the offer
+    to distribute corresponding source code.  (This alternative is
+    allowed only for noncommercial distribution and only if you
+    received the program in object code or executable form with such
+    an offer, in accord with Subsection b above.)
+
+The source code for a work means the preferred form of the work for
+making modifications to it.  For an executable work, complete source
+code means all the source code for all modules it contains, plus any
+associated interface definition files, plus the scripts used to
+control compilation and installation of the executable.  However, as a
+special exception, the source code distributed need not include
+anything that is normally distributed (in either source or binary
+form) with the major components (compiler, kernel, and so on) of the
+operating system on which the executable runs, unless that component
+itself accompanies the executable.
+
+If distribution of executable or object code is made by offering
+access to copy from a designated place, then offering equivalent
+access to copy the source code from the same place counts as
+distribution of the source code, even though third parties are not
+compelled to copy the source along with the object code.
+
+  4. You may not copy, modify, sublicense, or distribute the Program
+except as expressly provided under this License.  Any attempt
+otherwise to copy, modify, sublicense or distribute the Program is
+void, and will automatically terminate your rights under this License.
+However, parties who have received copies, or rights, from you under
+this License will not have their licenses terminated so long as such
+parties remain in full compliance.
+
+  5. You are not required to accept this License, since you have not
+signed it.  However, nothing else grants you permission to modify or
+distribute the Program or its derivative works.  These actions are
+prohibited by law if you do not accept this License.  Therefore, by
+modifying or distributing the Program (or any work based on the
+Program), you indicate your acceptance of this License to do so, and
+all its terms and conditions for copying, distributing or modifying
+the Program or works based on it.
+
+  6. Each time you redistribute the Program (or any work based on the
+Program), the recipient automatically receives a license from the
+original licensor to copy, distribute or modify the Program subject to
+these terms and conditions.  You may not impose any further
+restrictions on the recipients' exercise of the rights granted herein.
+You are not responsible for enforcing compliance by third parties to
+this License.
+
+  7. If, as a consequence of a court judgment or allegation of patent
+infringement or for any other reason (not limited to patent issues),
+conditions are imposed on you (whether by court order, agreement or
+otherwise) that contradict the conditions of this License, they do not
+excuse you from the conditions of this License.  If you cannot
+distribute so as to satisfy simultaneously your obligations under this
+License and any other pertinent obligations, then as a consequence you
+may not distribute the Program at all.  For example, if a patent
+license would not permit royalty-free redistribution of the Program by
+all those who receive copies directly or indirectly through you, then
+the only way you could satisfy both it and this License would be to
+refrain entirely from distribution of the Program.
+
+If any portion of this section is held invalid or unenforceable under
+any particular circumstance, the balance of the section is intended to
+apply and the section as a whole is intended to apply in other
+circumstances.
+
+It is not the purpose of this section to induce you to infringe any
+patents or other property right claims or to contest validity of any
+such claims; this section has the sole purpose of protecting the
+integrity of the free software distribution system, which is
+implemented by public license practices.  Many people have made
+generous contributions to the wide range of software distributed
+through that system in reliance on consistent application of that
+system; it is up to the author/donor to decide if he or she is willing
+to distribute software through any other system and a licensee cannot
+impose that choice.
+
+This section is intended to make thoroughly clear what is believed to
+be a consequence of the rest of this License.
+
+  8. If the distribution and/or use of the Program is restricted in
+certain countries either by patents or by copyrighted interfaces, the
+original copyright holder who places the Program under this License
+may add an explicit geographical distribution limitation excluding
+those countries, so that distribution is permitted only in or among
+countries not thus excluded.  In such case, this License incorporates
+the limitation as if written in the body of this License.
+
+  9. The Free Software Foundation may publish revised and/or new versions
+of the General Public License from time to time.  Such new versions will
+be similar in spirit to the present version, but may differ in detail to
+address new problems or concerns.
+
+Each version is given a distinguishing version number.  If the Program
+specifies a version number of this License which applies to it and "any
+later version", you have the option of following the terms and conditions
+either of that version or of any later version published by the Free
+Software Foundation.  If the Program does not specify a version number of
+this License, you may choose any version ever published by the Free Software
+Foundation.
+
+  10. If you wish to incorporate parts of the Program into other free
+programs whose distribution conditions are different, write to the author
+to ask for permission.  For software which is copyrighted by the Free
+Software Foundation, write to the Free Software Foundation; we sometimes
+make exceptions for this.  Our decision will be guided by the two goals
+of preserving the free status of all derivatives of our free software and
+of promoting the sharing and reuse of software generally.
+
+                            NO WARRANTY
+
+  11. BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
+FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW.  EXCEPT WHEN
+OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
+PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
+OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
+MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.  THE ENTIRE RISK AS
+TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU.  SHOULD THE
+PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING,
+REPAIR OR CORRECTION.
+
+  12. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
+WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
+REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
+INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING
+OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED
+TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY
+YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
+PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE
+POSSIBILITY OF SUCH DAMAGES.
+
+                     END OF TERMS AND CONDITIONS
+
+            How to Apply These Terms to Your New Programs
+
+  If you develop a new program, and you want it to be of the greatest
+possible use to the public, the best way to achieve this is to make it
+free software which everyone can redistribute and change under these terms.
+
+  To do so, attach the following notices to the program.  It is safest
+to attach them to the start of each source file to most effectively
+convey the exclusion of warranty; and each file should have at least
+the "copyright" line and a pointer to where the full notice is found.
+
+    <one line to give the program's name and a brief idea of what it does.>
+    Copyright (C) <year>  <name of author>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+Also add information on how to contact you by electronic and paper mail.
+
+If the program is interactive, make it output a short notice like this
+when it starts in an interactive mode:
+
+    Gnomovision version 69, Copyright (C) year name of author
+    Gnomovision comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
+    This is free software, and you are welcome to redistribute it
+    under certain conditions; type `show c' for details.
+
+The hypothetical commands `show w' and `show c' should show the appropriate
+parts of the General Public License.  Of course, the commands you use may
+be called something other than `show w' and `show c'; they could even be
+mouse-clicks or menu items--whatever suits your program.
+
+You should also get your employer (if you work as a programmer) or your
+school, if any, to sign a "copyright disclaimer" for the program, if
+necessary.  Here is a sample; alter the names:
+
+  Yoyodyne, Inc., hereby disclaims all copyright interest in the program
+  `Gnomovision' (which makes passes at compilers) written by James Hacker.
+
+  <signature of Ty Coon>, 1 April 1989
+  Ty Coon, President of Vice
+
+This General Public License does not permit incorporating your program into
+proprietary programs.  If your program is a subroutine library, you may
+consider it more useful to permit linking proprietary applications with the
+library.  If this is what you want to do, use the GNU Lesser General
+Public License instead of this License.

COPYING.UBDL

@@ -0,0 +1,59 @@
+UNMODIFIED BINARY DISTRIBUTION LICENCE
+
+
+PREAMBLE
+
+The GNU General Public License provides a legal guarantee that
+software covered by it remains free (in the sense of freedom, not
+price).  It achieves this guarantee by imposing obligations on anyone
+who chooses to distribute the software.
+
+Some of these obligations may be seen as unnecessarily burdensome.  In
+particular, when the source code for the software is already publicly
+and freely available, there is minimal value in imposing upon each
+distributor the obligation to provide the complete source code (or an
+equivalent written offer to provide the complete source code).
+
+This Licence allows for the distribution of unmodified binaries built
+from publicly available source code, without imposing the obligations
+of the GNU General Public License upon anyone who chooses to
+distribute only the unmodified binaries built from that source code.
+
+The extra permissions granted by this Licence apply only to unmodified
+binaries built from source code which has already been made available
+to the public in accordance with the terms of the GNU General Public
+Licence.  Nothing in this Licence allows for the creation of
+closed-source modified versions of the Program.  Any modified versions
+of the Program are subject to the usual terms and conditions of the
+GNU General Public License.
+
+
+TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
+
+This Licence applies to any Program or other work which contains a
+notice placed by the copyright holder saying it may be distributed
+under the terms of this Unmodified Binary Distribution Licence.  All
+terms used in the text of this Licence are to be interpreted as they
+are used in version 2 of the GNU General Public License as published
+by the Free Software Foundation.
+
+If you have made this Program available to the public in both source
+code and executable form in accordance with the terms of the GNU
+General Public License as published by the Free Software Foundation;
+either version 2 of the License, or (at your option) any later
+version, then you are hereby granted an additional permission to use,
+copy, and distribute the unmodified executable form of this Program
+(the "Unmodified Binary") without restriction, including the right to
+permit persons to whom the Unmodified Binary is furnished to do
+likewise, subject to the following conditions:
+
+- when started running, the Program must display an announcement which
+  includes the details of your existing publication of the Program
+  made in accordance with the terms of the GNU General Public License.
+  For example, the Program could display the URL of the publicly
+  available source code from which the Unmodified Binary was built.
+
+- when exercising your right to grant permissions under this Licence,
+  you do not need to refer directly to the text of this Licence, but
+  you may not grant permissions beyond those granted to you by this
+  Licence.

contrib/coverity/model.c

@@ -0,0 +1,29 @@
+/*
+ * Coverity modelling file
+ *
+ */
+
+typedef long off_t;
+typedef void * userptr_t;
+typedef long long time_t;
+struct tm;
+typedef unsigned short wchar_t;
+typedef void mbstate_t;
+struct digest_algorithm;
+
+/* Inhibit use of built-in models for functions where Coverity's
+ * assumptions about the modelled function are incorrect for iPXE.
+ */
+char * strerror ( int errno ) {
+}
+void copy_from_user ( void *dest, userptr_t src, off_t src_off, size_t len ) {
+}
+time_t mktime ( struct tm *tm ) {
+}
+int getchar ( void ) {
+}
+size_t wcrtomb ( char *buf, wchar_t wc, mbstate_t *ps ) {
+}
+void hmac_init ( struct digest_algorithm *digest, void *digest_ctx,
+		 void *key, size_t *key_len ) {
+}

src/Makefile.efi

@@ -0,0 +1,54 @@
+# -*- makefile -*- : Force emacs to use Makefile mode
+
+# Enable stack protection if available
+#
+SPG_TEST = $(CC) -fstack-protector-strong -mstack-protector-guard=global \
+		 -x c -c /dev/null -o /dev/null >/dev/null 2>&1
+SPG_FLAGS := $(shell $(SPG_TEST) && $(ECHO) '-fstack-protector-strong ' \
+					    '-mstack-protector-guard=global')
+CFLAGS += $(SPG_FLAGS)
+
+# The EFI linker script
+#
+LDSCRIPT	= scripts/efi.lds
+
+# Retain relocation information for elf2efi
+#
+LDFLAGS		+= -q -S
+
+# Media types.
+#
+NON_AUTO_MEDIA	+= efi
+NON_AUTO_MEDIA	+= efidrv
+NON_AUTO_MEDIA	+= drv.efi
+NON_AUTO_MEDIA	+= efirom
+
+# Include SNP driver in the all-drivers build
+#
+DRIVERS_net += snp
+
+# Rules for building EFI files
+#
+$(BIN)/%.efi : $(BIN)/%.efi.tmp $(ELF2EFI)
+	$(QM)$(ECHO) "  [FINISH] $@"
+	$(Q)$(ELF2EFI) --subsystem=10 $< $@
+
+$(BIN)/%.efidrv : $(BIN)/%.efidrv.tmp $(ELF2EFI)
+	$(QM)$(ECHO) "  [FINISH] $@"
+	$(Q)$(ELF2EFI) --subsystem=11 $< $@
+
+$(BIN)/%.drv.efi : $(BIN)/%.efidrv
+	$(QM)$(ECHO) "  [FINISH] $@"
+	$(Q)$(CP) $< $@
+
+$(BIN)/%.efirom : $(BIN)/%.efidrv $(EFIROM)
+	$(QM)$(ECHO) "  [FINISH] $@"
+	$(Q)$(EFIROM) -v $(TGT_PCI_VENDOR) -d $(TGT_PCI_DEVICE) $< $@
+
+$(BIN)/efidrv.cab : $(BIN)/alldrv.efis # $(ALL_drv.efi) is not yet defined
+	$(QM)$(ECHO) "  [CAB] $@"
+	$(Q)$(LCAB) -n -q $(ALL_drv.efi) $@
+
+$(BIN)/%.usb : $(BIN)/%.efi
+	$(QM)$(ECHO) "  [GENEFIDSK] $@"
+	$(Q)bash util/genefidsk -o $@ -b $(EFI_BOOT_FILE) $<

src/arch/arm/Makefile

@@ -0,0 +1,12 @@
+# Assembler section type character
+#
+ASM_TCHAR	:= %
+ASM_TCHAR_OPS	:= %%
+
+# Include common ARM headers
+#
+INCDIRS		+= arch/arm/include
+
+# ARM-specific directories containing source files
+#
+SRCDIRS		+= arch/arm/interface/efi

src/arch/arm/Makefile.efi

@@ -0,0 +1,6 @@
+# -*- makefile -*- : Force emacs to use Makefile mode
+
+# Include generic EFI Makefile
+#
+MAKEDEPS	+= Makefile.efi
+include Makefile.efi

src/arch/arm/core/arm_io.c

@@ -0,0 +1,93 @@
+/*
+ * Copyright (C) 2016 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <ipxe/io.h>
+#include <ipxe/arm_io.h>
+
+/** @file
+ *
+ * iPXE I/O API for ARM
+ *
+ */
+
+/** An ARM I/O qword */
+union arm32_io_qword {
+	uint64_t qword;
+	uint32_t dword[2];
+};
+
+/**
+ * Read 64-bit qword from memory-mapped device
+ *
+ * @v io_addr		I/O address
+ * @ret data		Value read
+ *
+ * This is not atomic for ARM32.
+ */
+static uint64_t arm32_readq ( volatile uint64_t *io_addr ) {
+	volatile union arm32_io_qword *ptr =
+		container_of ( io_addr, union arm32_io_qword, qword );
+	union arm32_io_qword tmp;
+
+	tmp.dword[0] = readl ( &ptr->dword[0] );
+	tmp.dword[1] = readl ( &ptr->dword[1] );
+	return tmp.qword;
+}
+
+/**
+ * Write 64-bit qword to memory-mapped device
+ *
+ * @v data		Value to write
+ * @v io_addr		I/O address
+ *
+ * This is not atomic for ARM32.
+ */
+static void arm32_writeq ( uint64_t data, volatile uint64_t *io_addr ) {
+	volatile union arm32_io_qword *ptr =
+		container_of ( io_addr, union arm32_io_qword, qword );
+	union arm32_io_qword tmp;
+
+	tmp.qword = data;
+	writel ( tmp.dword[0], &ptr->dword[0] );
+	writel ( tmp.dword[1], &ptr->dword[1] );
+}
+
+PROVIDE_IOAPI_INLINE ( arm, phys_to_bus );
+PROVIDE_IOAPI_INLINE ( arm, bus_to_phys );
+PROVIDE_IOAPI_INLINE ( arm, readb );
+PROVIDE_IOAPI_INLINE ( arm, readw );
+PROVIDE_IOAPI_INLINE ( arm, readl );
+PROVIDE_IOAPI_INLINE ( arm, writeb );
+PROVIDE_IOAPI_INLINE ( arm, writew );
+PROVIDE_IOAPI_INLINE ( arm, writel );
+PROVIDE_IOAPI_INLINE ( arm, iodelay );
+PROVIDE_IOAPI_INLINE ( arm, mb );
+#ifdef __aarch64__
+PROVIDE_IOAPI_INLINE ( arm, readq );
+PROVIDE_IOAPI_INLINE ( arm, writeq );
+#else
+PROVIDE_IOAPI ( arm, readq, arm32_readq );
+PROVIDE_IOAPI ( arm, writeq, arm32_writeq );
+#endif

src/arch/arm/include/bits/acpi.h

@@ -0,0 +1,12 @@
+#ifndef _BITS_ACPI_H
+#define _BITS_ACPI_H
+
+/** @file
+ *
+ * ARM-specific ACPI API implementations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#endif /* _BITS_ACPI_H */

src/arch/arm/include/bits/endian.h

@@ -0,0 +1,13 @@
+#ifndef _BITS_ENDIAN_H
+#define _BITS_ENDIAN_H
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/* ARM may be either little-endian or big-endian */
+#ifdef __ARM_BIG_ENDIAN
+#define __BYTE_ORDER __BIG_ENDIAN
+#else
+#define __BYTE_ORDER __LITTLE_ENDIAN
+#endif
+
+#endif /* _BITS_ENDIAN_H */

src/arch/arm/include/bits/entropy.h

@@ -0,0 +1,12 @@
+#ifndef _BITS_ENTROPY_H
+#define _BITS_ENTROPY_H
+
+/** @file
+ *
+ * ARM-specific entropy API implementations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#endif /* _BITS_ENTROPY_H */

src/arch/arm/include/bits/errfile.h

@@ -0,0 +1,19 @@
+#ifndef _BITS_ERRFILE_H
+#define _BITS_ERRFILE_H
+
+/** @file
+ *
+ * ARM-specific error file identifiers
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/**
+ * @addtogroup errfile Error file identifiers
+ * @{
+ */
+
+/** @} */
+
+#endif /* _BITS_ERRFILE_H */

src/arch/arm/include/bits/hyperv.h

@@ -0,0 +1,12 @@
+#ifndef _BITS_HYPERV_H
+#define _BITS_HYPERV_H
+
+/** @file
+ *
+ * Hyper-V interface
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#endif /* _BITS_HYPERV_H */

src/arch/arm/include/bits/io.h

@@ -0,0 +1,14 @@
+#ifndef _BITS_IO_H
+#define _BITS_IO_H
+
+/** @file
+ *
+ * ARM-specific I/O API implementations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <ipxe/arm_io.h>
+
+#endif /* _BITS_IO_H */

src/arch/arm/include/bits/iomap.h

@@ -0,0 +1,12 @@
+#ifndef _BITS_IOMAP_H
+#define _BITS_IOMAP_H
+
+/** @file
+ *
+ * ARM-specific I/O mapping API implementations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#endif /* _BITS_IOMAP_H */

src/arch/arm/include/bits/nap.h

@@ -0,0 +1,14 @@
+#ifndef _BITS_NAP_H
+#define _BITS_NAP_H
+
+/** @file
+ *
+ * ARM-specific CPU sleeping API implementations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <ipxe/efi/efiarm_nap.h>
+
+#endif /* _BITS_MAP_H */

src/arch/arm/include/bits/pci_io.h

@@ -0,0 +1,14 @@
+#ifndef _BITS_PCI_IO_H
+#define _BITS_PCI_IO_H
+
+/** @file
+ *
+ * ARM PCI I/O API implementations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <ipxe/io.h>
+
+#endif /* _BITS_PCI_IO_H */

src/arch/arm/include/bits/reboot.h

@@ -0,0 +1,12 @@
+#ifndef _BITS_REBOOT_H
+#define _BITS_REBOOT_H
+
+/** @file
+ *
+ * ARM-specific reboot API implementations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#endif /* _BITS_REBOOT_H */

src/arch/arm/include/bits/sanboot.h

@@ -0,0 +1,12 @@
+#ifndef _BITS_SANBOOT_H
+#define _BITS_SANBOOT_H
+
+/** @file
+ *
+ * ARM-specific sanboot API implementations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#endif /* _BITS_SANBOOT_H */

src/arch/arm/include/bits/smbios.h

@@ -0,0 +1,12 @@
+#ifndef _BITS_SMBIOS_H
+#define _BITS_SMBIOS_H
+
+/** @file
+ *
+ * ARM-specific SMBIOS API implementations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#endif /* _BITS_SMBIOS_H */

src/arch/arm/include/bits/time.h

@@ -0,0 +1,12 @@
+#ifndef _BITS_TIME_H
+#define _BITS_TIME_H
+
+/** @file
+ *
+ * ARM-specific time API implementations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#endif /* _BITS_TIME_H */

src/arch/arm/include/bits/uaccess.h

@@ -0,0 +1,12 @@
+#ifndef _BITS_UACCESS_H
+#define _BITS_UACCESS_H
+
+/** @file
+ *
+ * ARM-specific user access API implementations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#endif /* _BITS_UACCESS_H */

src/arch/arm/include/bits/uart.h

@@ -0,0 +1,12 @@
+#ifndef _BITS_UART_H
+#define _BITS_UART_H
+
+/** @file
+ *
+ * 16550-compatible UART
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#endif /* _BITS_UART_H */

src/arch/arm/include/bits/umalloc.h

@@ -0,0 +1,12 @@
+#ifndef _BITS_UMALLOC_H
+#define _BITS_UMALLOC_H
+
+/** @file
+ *
+ * ARM-specific user memory allocation API implementations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#endif /* _BITS_UMALLOC_H */

src/arch/arm/include/bits/xen.h

@@ -0,0 +1,158 @@
+#ifndef _BITS_XEN_H
+#define _BITS_XEN_H
+
+/** @file
+ *
+ * Xen interface
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/* Hypercall registers */
+#ifdef __aarch64__
+#define XEN_HC "x16"
+#define XEN_REG1 "x0"
+#define XEN_REG2 "x1"
+#define XEN_REG3 "x2"
+#define XEN_REG4 "x3"
+#define XEN_REG5 "x4"
+#else
+#define XEN_HC "r12"
+#define XEN_REG1 "r0"
+#define XEN_REG2 "r1"
+#define XEN_REG3 "r2"
+#define XEN_REG4 "r3"
+#define XEN_REG5 "r4"
+#endif
+
+/**
+ * Issue hypercall with one argument
+ *
+ * @v xen		Xen hypervisor
+ * @v hypercall		Hypercall number
+ * @v arg1		First argument
+ * @ret retval		Return value
+ */
+static inline __attribute__ (( always_inline )) unsigned long
+xen_hypercall_1 ( struct xen_hypervisor *xen __unused, unsigned int hypercall,
+		  unsigned long arg1 ) {
+	register unsigned long hc asm ( XEN_HC ) = hypercall;
+	register unsigned long reg1 asm ( XEN_REG1 ) = arg1;
+
+	__asm__ __volatile__ ( "hvc %1"
+			       : "+r" ( reg1 )
+			       : "i" ( XEN_HYPERCALL_TAG ), "r" ( hc )
+			       : "memory", "cc" );
+	return reg1;
+}
+
+/**
+ * Issue hypercall with two arguments
+ *
+ * @v xen		Xen hypervisor
+ * @v hypercall		Hypercall number
+ * @v arg1		First argument
+ * @v arg2		Second argument
+ * @ret retval		Return value
+ */
+static inline __attribute__ (( always_inline )) unsigned long
+xen_hypercall_2 ( struct xen_hypervisor *xen __unused, unsigned int hypercall,
+		  unsigned long arg1, unsigned long arg2 ) {
+	register unsigned long hc asm ( XEN_HC ) = hypercall;
+	register unsigned long reg1 asm ( XEN_REG1 ) = arg1;
+	register unsigned long reg2 asm ( XEN_REG2 ) = arg2;
+
+	__asm__ __volatile__ ( "hvc %2"
+			       : "+r" ( reg1 ), "+r" ( reg2 )
+			       : "i" ( XEN_HYPERCALL_TAG ), "r" ( hc )
+			       : "memory", "cc" );
+	return reg1;
+}
+
+/**
+ * Issue hypercall with three arguments
+ *
+ * @v xen		Xen hypervisor
+ * @v hypercall		Hypercall number
+ * @v arg1		First argument
+ * @v arg2		Second argument
+ * @v arg3		Third argument
+ * @ret retval		Return value
+ */
+static inline __attribute__ (( always_inline )) unsigned long
+xen_hypercall_3 ( struct xen_hypervisor *xen __unused, unsigned int hypercall,
+		  unsigned long arg1, unsigned long arg2, unsigned long arg3 ) {
+	register unsigned long hc asm ( XEN_HC ) = hypercall;
+	register unsigned long reg1 asm ( XEN_REG1 ) = arg1;
+	register unsigned long reg2 asm ( XEN_REG2 ) = arg2;
+	register unsigned long reg3 asm ( XEN_REG3 ) = arg3;
+
+	__asm__ __volatile__ ( "hvc %3"
+			       : "+r" ( reg1 ), "+r" ( reg2 ), "+r" ( reg3 )
+			       : "i" ( XEN_HYPERCALL_TAG ), "r" ( hc )
+			       : "memory", "cc" );
+	return reg1;
+}
+
+/**
+ * Issue hypercall with four arguments
+ *
+ * @v xen		Xen hypervisor
+ * @v hypercall		Hypercall number
+ * @v arg1		First argument
+ * @v arg2		Second argument
+ * @v arg3		Third argument
+ * @v arg4		Fourth argument
+ * @ret retval		Return value
+ */
+static inline __attribute__ (( always_inline )) unsigned long
+xen_hypercall_4 ( struct xen_hypervisor *xen __unused, unsigned int hypercall,
+		  unsigned long arg1, unsigned long arg2, unsigned long arg3,
+		  unsigned long arg4 ) {
+	register unsigned long hc asm ( XEN_HC ) = hypercall;
+	register unsigned long reg1 asm ( XEN_REG1 ) = arg1;
+	register unsigned long reg2 asm ( XEN_REG2 ) = arg2;
+	register unsigned long reg3 asm ( XEN_REG3 ) = arg3;
+	register unsigned long reg4 asm ( XEN_REG4 ) = arg4;
+
+	__asm__ __volatile__ ( "hvc %4"
+			       : "+r" ( reg1 ), "+r" ( reg2 ), "+r" ( reg3 ),
+				 "+r" ( reg4 )
+			       : "i" ( XEN_HYPERCALL_TAG ), "r" ( hc )
+			       : "memory", "cc" );
+	return reg1;
+}
+
+/**
+ * Issue hypercall with five arguments
+ *
+ * @v xen		Xen hypervisor
+ * @v hypercall		Hypercall number
+ * @v arg1		First argument
+ * @v arg2		Second argument
+ * @v arg3		Third argument
+ * @v arg4		Fourth argument
+ * @v arg5		Fifth argument
+ * @ret retval		Return value
+ */
+static inline __attribute__ (( always_inline )) unsigned long
+xen_hypercall_5 ( struct xen_hypervisor *xen __unused, unsigned int hypercall,
+		  unsigned long arg1, unsigned long arg2, unsigned long arg3,
+		  unsigned long arg4, unsigned long arg5 ) {
+	register unsigned long hc asm ( XEN_HC ) = hypercall;
+	register unsigned long reg1 asm ( XEN_REG1 ) = arg1;
+	register unsigned long reg2 asm ( XEN_REG2 ) = arg2;
+	register unsigned long reg3 asm ( XEN_REG3 ) = arg3;
+	register unsigned long reg4 asm ( XEN_REG4 ) = arg4;
+	register unsigned long reg5 asm ( XEN_REG5 ) = arg5;
+
+	__asm__ __volatile__ ( "hvc %5"
+			       : "+r" ( reg1 ), "+r" ( reg2 ), "+r" ( reg3 ),
+				 "+r" ( reg4 ), "+r" ( reg5 )
+			       : "i" ( XEN_HYPERCALL_TAG ), "r" ( hc )
+			       : "memory", "cc" );
+	return reg1;
+}
+
+#endif /* _BITS_XEN_H */

src/arch/arm/include/ipxe/arm_io.h

@@ -0,0 +1,146 @@
+#ifndef _IPXE_ARM_IO_H
+#define _IPXE_ARM_IO_H
+
+/** @file
+ *
+ * iPXE I/O API for ARM
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#ifdef IOAPI_ARM
+#define IOAPI_PREFIX_arm
+#else
+#define IOAPI_PREFIX_arm __arm_
+#endif
+
+/*
+ * Memory space mappings
+ *
+ */
+
+/** Page shift */
+#define PAGE_SHIFT 12
+
+/*
+ * Physical<->Bus address mappings
+ *
+ */
+
+static inline __always_inline unsigned long
+IOAPI_INLINE ( arm, phys_to_bus ) ( unsigned long phys_addr ) {
+	return phys_addr;
+}
+
+static inline __always_inline unsigned long
+IOAPI_INLINE ( arm, bus_to_phys ) ( unsigned long bus_addr ) {
+	return bus_addr;
+}
+
+/*
+ * MMIO reads and writes up to native word size
+ *
+ */
+
+#define ARM_READX( _suffix, _type, _insn_suffix, _reg_prefix )		      \
+static inline __always_inline _type					      \
+IOAPI_INLINE ( arm, read ## _suffix ) ( volatile _type *io_addr ) {	      \
+	_type data;							      \
+	__asm__ __volatile__ ( "ldr" _insn_suffix " %" _reg_prefix "0, %1"    \
+			       : "=r" ( data ) : "Qo" ( *io_addr ) );	      \
+	return data;							      \
+}
+#ifdef __aarch64__
+ARM_READX ( b, uint8_t, "b", "w" );
+ARM_READX ( w, uint16_t, "h", "w" );
+ARM_READX ( l, uint32_t, "", "w" );
+ARM_READX ( q, uint64_t, "", "" );
+#else
+ARM_READX ( b, uint8_t, "b", "" );
+ARM_READX ( w, uint16_t, "h", "" );
+ARM_READX ( l, uint32_t, "", "" );
+#endif
+
+#define ARM_WRITEX( _suffix, _type, _insn_suffix, _reg_prefix )		      \
+static inline __always_inline void					      \
+IOAPI_INLINE ( arm, write ## _suffix ) ( _type data,			      \
+					 volatile _type *io_addr ) {	      \
+	__asm__ __volatile__ ( "str" _insn_suffix " %" _reg_prefix "0, %1"    \
+			       : : "r" ( data ), "Qo" ( *io_addr ) );	      \
+}
+#ifdef __aarch64__
+ARM_WRITEX ( b, uint8_t, "b", "w" );
+ARM_WRITEX ( w, uint16_t, "h", "w" );
+ARM_WRITEX ( l, uint32_t, "", "w" );
+ARM_WRITEX ( q, uint64_t, "", "" );
+#else
+ARM_WRITEX ( b, uint8_t, "b", "" );
+ARM_WRITEX ( w, uint16_t, "h", "" );
+ARM_WRITEX ( l, uint32_t, "", "" );
+#endif
+
+/*
+ * Dummy PIO reads and writes up to 32 bits
+ *
+ * There is no common standard for I/O-space access for ARM, and
+ * non-MMIO peripherals are vanishingly rare.  Provide dummy
+ * implementations that will allow code to link and should cause
+ * drivers to simply fail to detect hardware at runtime.
+ *
+ */
+
+#define ARM_INX( _suffix, _type )					      \
+static inline __always_inline _type					      \
+IOAPI_INLINE ( arm, in ## _suffix ) ( volatile _type *io_addr __unused) {     \
+	return ~( (_type) 0 );						      \
+}									      \
+static inline __always_inline void					      \
+IOAPI_INLINE ( arm, ins ## _suffix ) ( volatile _type *io_addr __unused,      \
+				       _type *data, unsigned int count ) {    \
+	memset ( data, 0xff, count * sizeof ( *data ) );		      \
+}
+ARM_INX ( b, uint8_t );
+ARM_INX ( w, uint16_t );
+ARM_INX ( l, uint32_t );
+
+#define ARM_OUTX( _suffix, _type )					      \
+static inline __always_inline void					      \
+IOAPI_INLINE ( arm, out ## _suffix ) ( _type data __unused,		      \
+				       volatile _type *io_addr __unused ) {   \
+	/* Do nothing */						      \
+}									      \
+static inline __always_inline void					      \
+IOAPI_INLINE ( arm, outs ## _suffix ) ( volatile _type *io_addr __unused,     \
+					const _type *data __unused,	      \
+					unsigned int count __unused ) {	      \
+	/* Do nothing */						      \
+}
+ARM_OUTX ( b, uint8_t );
+ARM_OUTX ( w, uint16_t );
+ARM_OUTX ( l, uint32_t );
+
+/*
+ * Slow down I/O
+ *
+ */
+static inline __always_inline void
+IOAPI_INLINE ( arm, iodelay ) ( void ) {
+	/* Nothing to do */
+}
+
+/*
+ * Memory barrier
+ *
+ */
+static inline __always_inline void
+IOAPI_INLINE ( arm, mb ) ( void ) {
+
+#ifdef __aarch64__
+	__asm__ __volatile__ ( "dmb sy" );
+#else
+	__asm__ __volatile__ ( "dmb" );
+#endif
+}
+
+#endif /* _IPXE_ARM_IO_H */

src/arch/arm/include/ipxe/efi/efiarm_nap.h

@@ -0,0 +1,18 @@
+#ifndef _IPXE_EFIARM_NAP_H
+#define _IPXE_EFIARM_NAP_H
+
+/** @file
+ *
+ * EFI CPU sleeping
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#ifdef NAP_EFIARM
+#define NAP_PREFIX_efiarm
+#else
+#define NAP_PREFIX_efiarm __efiarm_
+#endif
+
+#endif /* _IPXE_EFIARM_NAP_H */

src/arch/arm/interface/efi/efiarm_nap.c

@@ -0,0 +1,53 @@
+/*
+ * Copyright (C) 2016 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <ipxe/nap.h>
+#include <ipxe/efi/efi.h>
+
+/** @file
+ *
+ * iPXE CPU sleeping API for EFI
+ *
+ */
+
+/**
+ * Sleep until next interrupt
+ *
+ */
+static void efiarm_cpu_nap ( void ) {
+	/*
+	 * I can't find any EFI API that allows us to put the CPU to
+	 * sleep.  The CpuSleep() function is defined in CpuLib.h, but
+	 * isn't part of any exposed protocol so we have no way to
+	 * call it.
+	 *
+	 * The EFI shell doesn't seem to bother sleeping the CPU; it
+	 * just sits there idly burning power.
+	 *
+	 */
+	__asm__ __volatile__ ( "wfi" );
+}
+
+PROVIDE_NAP ( efiarm, cpu_nap, efiarm_cpu_nap );

src/arch/arm32/Makefile

@@ -0,0 +1,23 @@
+# ARM32-specific directories containing source files
+#
+SRCDIRS		+= arch/arm32/core
+SRCDIRS		+= arch/arm32/libgcc
+
+# ARM32-specific flags
+#
+CFLAGS		+= -mthumb -mcpu=cortex-a15 -mabi=aapcs -mfloat-abi=soft
+CFLAGS		+= -mword-relocations
+ASFLAGS		+= -mthumb -mcpu=cortex-a15
+
+# EFI requires -fshort-wchar, and nothing else currently uses wchar_t
+#
+CFLAGS		+= -fshort-wchar
+
+# Include common ARM Makefile
+MAKEDEPS	+= arch/arm/Makefile
+include arch/arm/Makefile
+
+# Include platform-specific Makefile
+#
+MAKEDEPS	+= arch/arm32/Makefile.$(PLATFORM)
+include arch/arm32/Makefile.$(PLATFORM)

src/arch/arm32/Makefile.efi

@@ -0,0 +1,18 @@
+# -*- makefile -*- : Force emacs to use Makefile mode
+
+# UEFI requires that enums are always 32 bits
+#
+CFLAGS		+= -fno-short-enums
+
+# Specify EFI image builder
+#
+ELF2EFI		= $(ELF2EFI32)
+
+# Specify EFI boot file
+#
+EFI_BOOT_FILE	= bootarm.efi
+
+# Include generic EFI Makefile
+#
+MAKEDEPS	+= arch/arm/Makefile.efi
+include arch/arm/Makefile.efi

src/arch/arm32/core/arm32_bigint.c

@@ -0,0 +1,102 @@
+/*
+ * Copyright (C) 2016 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+#include <string.h>
+#include <ipxe/bigint.h>
+
+/** @file
+ *
+ * Big integer support
+ */
+
+/**
+ * Multiply big integers
+ *
+ * @v multiplicand0	Element 0 of big integer to be multiplied
+ * @v multiplier0	Element 0 of big integer to be multiplied
+ * @v result0		Element 0 of big integer to hold result
+ * @v size		Number of elements
+ */
+void bigint_multiply_raw ( const uint32_t *multiplicand0,
+			   const uint32_t *multiplier0,
+			   uint32_t *result0, unsigned int size ) {
+	const bigint_t ( size ) __attribute__ (( may_alias )) *multiplicand =
+		( ( const void * ) multiplicand0 );
+	const bigint_t ( size ) __attribute__ (( may_alias )) *multiplier =
+		( ( const void * ) multiplier0 );
+	bigint_t ( size * 2 ) __attribute__ (( may_alias )) *result =
+		( ( void * ) result0 );
+	unsigned int i;
+	unsigned int j;
+	uint32_t multiplicand_element;
+	uint32_t multiplier_element;
+	uint32_t *result_elements;
+	uint32_t discard_low;
+	uint32_t discard_high;
+	uint32_t discard_temp;
+
+	/* Zero result */
+	memset ( result, 0, sizeof ( *result ) );
+
+	/* Multiply integers one element at a time */
+	for ( i = 0 ; i < size ; i++ ) {
+		multiplicand_element = multiplicand->element[i];
+		for ( j = 0 ; j < size ; j++ ) {
+			multiplier_element = multiplier->element[j];
+			result_elements = &result->element[ i + j ];
+			/* Perform a single multiply, and add the
+			 * resulting double-element into the result,
+			 * carrying as necessary.  The carry can
+			 * never overflow beyond the end of the
+			 * result, since:
+			 *
+			 *     a < 2^{n}, b < 2^{n} => ab < 2^{2n}
+			 */
+			__asm__ __volatile__ ( "umull %1, %2, %5, %6\n\t"
+					       "ldr %3, [%0]\n\t"
+					       "adds %3, %1\n\t"
+					       "stmia %0!, {%3}\n\t"
+					       "ldr %3, [%0]\n\t"
+					       "adcs %3, %2\n\t"
+					       "stmia %0!, {%3}\n\t"
+					       "bcc 2f\n\t"
+					       "\n1:\n\t"
+					       "ldr %3, [%0]\n\t"
+					       "adcs %3, #0\n\t"
+					       "stmia %0!, {%3}\n\t"
+					       "bcs 1b\n\t"
+					       "\n2:\n\t"
+					       : "+l" ( result_elements ),
+						 "=l" ( discard_low ),
+						 "=l" ( discard_high ),
+						 "=l" ( discard_temp ),
+						 "+m" ( *result )
+					       : "l" ( multiplicand_element ),
+						 "l" ( multiplier_element )
+					       : "cc" );
+		}
+	}
+}

src/arch/arm32/core/setjmp.S

@@ -0,0 +1,32 @@
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL )
+
+	.text
+	.arm
+
+/*
+ * Save stack context for non-local goto
+ */
+	.globl	setjmp
+	.type	setjmp, %function
+setjmp:
+	/* Store registers */
+	stmia	r0, { r4, r5, r6, r7, r8, r9, r10, fp, sp, lr }
+	/* Return 0 when returning as setjmp() */
+	mov	r0, #0
+	bx	lr
+	.size	setjmp, . - setjmp
+
+/*
+ * Non-local jump to a saved stack context
+ */
+	.globl	longjmp
+	.type	longjmp, %function
+longjmp:
+	/* Restore registers */
+	ldmia	r0, { r4, r5, r6, r7, r8, r9, r10, fp, sp, lr }
+	/* Force result to non-zero */
+	movs	r0, r1
+	moveq	r0, #1
+	/* Return to setjmp() caller */
+	bx	lr
+	.size	longjmp, . - longjmp

src/arch/arm32/include/bits/bigint.h

@@ -0,0 +1,316 @@
+#ifndef _BITS_BIGINT_H
+#define _BITS_BIGINT_H
+
+/** @file
+ *
+ * Big integer support
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+#include <string.h>
+#include <strings.h>
+
+/** Element of a big integer */
+typedef uint32_t bigint_element_t;
+
+/**
+ * Initialise big integer
+ *
+ * @v value0		Element 0 of big integer to initialise
+ * @v size		Number of elements
+ * @v data		Raw data
+ * @v len		Length of raw data
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_init_raw ( uint32_t *value0, unsigned int size,
+		  const void *data, size_t len ) {
+	size_t pad_len = ( sizeof ( bigint_t ( size ) ) - len );
+	uint8_t *value_byte = ( ( void * ) value0 );
+	const uint8_t *data_byte = ( data + len );
+
+	/* Copy raw data in reverse order, padding with zeros */
+	while ( len-- )
+		*(value_byte++) = *(--data_byte);
+	while ( pad_len-- )
+		*(value_byte++) = 0;
+}
+
+/**
+ * Add big integers
+ *
+ * @v addend0		Element 0 of big integer to add
+ * @v value0		Element 0 of big integer to be added to
+ * @v size		Number of elements
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_add_raw ( const uint32_t *addend0, uint32_t *value0,
+		 unsigned int size ) {
+	bigint_t ( size ) __attribute__ (( may_alias )) *value =
+		( ( void * ) value0 );
+	uint32_t *discard_addend;
+	uint32_t *discard_value;
+	uint32_t *discard_end;
+	uint32_t discard_addend_i;
+	uint32_t discard_value_i;
+
+	__asm__ __volatile__ ( "adds %2, %0, %8, lsl #2\n\t" /* clear CF */
+			       "\n1:\n\t"
+			       "ldmia %0!, {%3}\n\t"
+			       "ldr %4, [%1]\n\t"
+			       "adcs %4, %3\n\t"
+			       "stmia %1!, {%4}\n\t"
+			       "teq %0, %2\n\t"
+			       "bne 1b\n\t"
+			       : "=l" ( discard_addend ),
+				 "=l" ( discard_value ),
+				 "=l" ( discard_end ),
+				 "=l" ( discard_addend_i ),
+				 "=l" ( discard_value_i ),
+				 "+m" ( *value )
+			       : "0" ( addend0 ), "1" ( value0 ), "l" ( size )
+			       : "cc" );
+}
+
+/**
+ * Subtract big integers
+ *
+ * @v subtrahend0	Element 0 of big integer to subtract
+ * @v value0		Element 0 of big integer to be subtracted from
+ * @v size		Number of elements
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_subtract_raw ( const uint32_t *subtrahend0, uint32_t *value0,
+		      unsigned int size ) {
+	bigint_t ( size ) __attribute__ (( may_alias )) *value =
+		( ( void * ) value0 );
+	uint32_t *discard_subtrahend;
+	uint32_t *discard_value;
+	uint32_t *discard_end;
+	uint32_t discard_subtrahend_i;
+	uint32_t discard_value_i;
+
+	__asm__ __volatile__ ( "add %2, %0, %8, lsl #2\n\t"
+			       "cmp %2, %0\n\t" /* set CF */
+			       "\n1:\n\t"
+			       "ldmia %0!, {%3}\n\t"
+			       "ldr %4, [%1]\n\t"
+			       "sbcs %4, %3\n\t"
+			       "stmia %1!, {%4}\n\t"
+			       "teq %0, %2\n\t"
+			       "bne 1b\n\t"
+			       : "=l" ( discard_subtrahend ),
+				 "=l" ( discard_value ),
+				 "=l" ( discard_end ),
+				 "=l" ( discard_subtrahend_i ),
+				 "=l" ( discard_value_i ),
+				 "+m" ( *value )
+			       : "0" ( subtrahend0 ), "1" ( value0 ),
+				 "l" ( size )
+			       : "cc" );
+}
+
+/**
+ * Rotate big integer left
+ *
+ * @v value0		Element 0 of big integer
+ * @v size		Number of elements
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_rol_raw ( uint32_t *value0, unsigned int size ) {
+	bigint_t ( size ) __attribute__ (( may_alias )) *value =
+		( ( void * ) value0 );
+	uint32_t *discard_value;
+	uint32_t *discard_end;
+	uint32_t discard_value_i;
+
+	__asm__ __volatile__ ( "adds %1, %0, %5, lsl #2\n\t" /* clear CF */
+			       "\n1:\n\t"
+			       "ldr %2, [%0]\n\t"
+			       "adcs %2, %2\n\t"
+			       "stmia %0!, {%2}\n\t"
+			       "teq %0, %1\n\t"
+			       "bne 1b\n\t"
+			       : "=l" ( discard_value ),
+				 "=l" ( discard_end ),
+				 "=l" ( discard_value_i ),
+				 "+m" ( *value )
+			       : "0" ( value0 ), "1" ( size )
+			       : "cc" );
+}
+
+/**
+ * Rotate big integer right
+ *
+ * @v value0		Element 0 of big integer
+ * @v size		Number of elements
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_ror_raw ( uint32_t *value0, unsigned int size ) {
+	bigint_t ( size ) __attribute__ (( may_alias )) *value =
+		( ( void * ) value0 );
+	uint32_t *discard_value;
+	uint32_t *discard_end;
+	uint32_t discard_value_i;
+
+	__asm__ __volatile__ ( "adds %1, %0, %5, lsl #2\n\t" /* clear CF */
+			       "\n1:\n\t"
+			       "ldmdb %1!, {%2}\n\t"
+			       "rrxs %2, %2\n\t"
+			       "str %2, [%1]\n\t"
+			       "teq %0, %1\n\t"
+			       "bne 1b\n\t"
+			       : "=l" ( discard_value ),
+				 "=l" ( discard_end ),
+				 "=l" ( discard_value_i ),
+				 "+m" ( *value )
+			       : "0" ( value0 ), "1" ( size )
+			       : "cc" );
+}
+
+/**
+ * Test if big integer is equal to zero
+ *
+ * @v value0		Element 0 of big integer
+ * @v size		Number of elements
+ * @ret is_zero		Big integer is equal to zero
+ */
+static inline __attribute__ (( always_inline, pure )) int
+bigint_is_zero_raw ( const uint32_t *value0, unsigned int size ) {
+	const uint32_t *value = value0;
+	uint32_t value_i;
+
+	do {
+		value_i = *(value++);
+		if ( value_i )
+			break;
+	} while ( --size );
+
+	return ( value_i == 0 );
+}
+
+/**
+ * Compare big integers
+ *
+ * @v value0		Element 0 of big integer
+ * @v reference0	Element 0 of reference big integer
+ * @v size		Number of elements
+ * @ret geq		Big integer is greater than or equal to the reference
+ */
+static inline __attribute__ (( always_inline, pure )) int
+bigint_is_geq_raw ( const uint32_t *value0, const uint32_t *reference0,
+		    unsigned int size ) {
+	const uint32_t *value = ( value0 + size );
+	const uint32_t *reference = ( reference0 + size );
+	uint32_t value_i;
+	uint32_t reference_i;
+
+	do {
+		value_i = *(--value);
+		reference_i = *(--reference);
+		if ( value_i != reference_i )
+			break;
+	} while ( --size );
+
+	return ( value_i >= reference_i );
+}
+
+/**
+ * Test if bit is set in big integer
+ *
+ * @v value0		Element 0 of big integer
+ * @v size		Number of elements
+ * @v bit		Bit to test
+ * @ret is_set		Bit is set
+ */
+static inline __attribute__ (( always_inline )) int
+bigint_bit_is_set_raw ( const uint32_t *value0, unsigned int size,
+			unsigned int bit ) {
+	const bigint_t ( size ) __attribute__ (( may_alias )) *value =
+		( ( const void * ) value0 );
+	unsigned int index = ( bit / ( 8 * sizeof ( value->element[0] ) ) );
+	unsigned int subindex = ( bit % ( 8 * sizeof ( value->element[0] ) ) );
+
+	return ( value->element[index] & ( 1 << subindex ) );
+}
+
+/**
+ * Find highest bit set in big integer
+ *
+ * @v value0		Element 0 of big integer
+ * @v size		Number of elements
+ * @ret max_bit		Highest bit set + 1 (or 0 if no bits set)
+ */
+static inline __attribute__ (( always_inline )) int
+bigint_max_set_bit_raw ( const uint32_t *value0, unsigned int size ) {
+	const uint32_t *value = ( value0 + size );
+	int max_bit = ( 8 * sizeof ( bigint_t ( size ) ) );
+	uint32_t value_i;
+
+	do {
+		value_i = *(--value);
+		max_bit -= ( 32 - fls ( value_i ) );
+		if ( value_i )
+			break;
+	} while ( --size );
+
+	return max_bit;
+}
+
+/**
+ * Grow big integer
+ *
+ * @v source0		Element 0 of source big integer
+ * @v source_size	Number of elements in source big integer
+ * @v dest0		Element 0 of destination big integer
+ * @v dest_size		Number of elements in destination big integer
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_grow_raw ( const uint32_t *source0, unsigned int source_size,
+		  uint32_t *dest0, unsigned int dest_size ) {
+	unsigned int pad_size = ( dest_size - source_size );
+
+	memcpy ( dest0, source0, sizeof ( bigint_t ( source_size ) ) );
+	memset ( ( dest0 + source_size ), 0, sizeof ( bigint_t ( pad_size ) ) );
+}
+
+/**
+ * Shrink big integer
+ *
+ * @v source0		Element 0 of source big integer
+ * @v source_size	Number of elements in source big integer
+ * @v dest0		Element 0 of destination big integer
+ * @v dest_size		Number of elements in destination big integer
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_shrink_raw ( const uint32_t *source0, unsigned int source_size __unused,
+		    uint32_t *dest0, unsigned int dest_size ) {
+
+	memcpy ( dest0, source0, sizeof ( bigint_t ( dest_size ) ) );
+}
+
+/**
+ * Finalise big integer
+ *
+ * @v value0		Element 0 of big integer to finalise
+ * @v size		Number of elements
+ * @v out		Output buffer
+ * @v len		Length of output buffer
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_done_raw ( const uint32_t *value0, unsigned int size __unused,
+		  void *out, size_t len ) {
+	const uint8_t *value_byte = ( ( const void * ) value0 );
+	uint8_t *out_byte = ( out + len );
+
+	/* Copy raw data in reverse order */
+	while ( len-- )
+		*(--out_byte) = *(value_byte++);
+}
+
+extern void bigint_multiply_raw ( const uint32_t *multiplicand0,
+				  const uint32_t *multiplier0,
+				  uint32_t *value0, unsigned int size );
+
+#endif /* _BITS_BIGINT_H */

src/arch/arm32/include/bits/bitops.h

@@ -0,0 +1,100 @@
+#ifndef _BITS_BITOPS_H
+#define _BITS_BITOPS_H
+
+/** @file
+ *
+ * ARM bit operations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+
+/**
+ * Test and set bit atomically
+ *
+ * @v bit		Bit to set
+ * @v bits		Bit field
+ * @ret old		Old value of bit (zero or non-zero)
+ */
+static inline __attribute__ (( always_inline )) int
+test_and_set_bit ( unsigned int bit, volatile void *bits ) {
+	unsigned int index = ( bit / 32 );
+	unsigned int offset = ( bit % 32 );
+	volatile uint32_t *dword = ( ( ( volatile uint32_t * ) bits ) + index );
+	uint32_t mask = ( 1UL << offset );
+	uint32_t old;
+	uint32_t new;
+	uint32_t flag;
+
+	__asm__ __volatile__ ( "\n1:\n\t"
+			       "ldrex %0, %3\n\t"
+			       "orr %1, %0, %4\n\t"
+			       "strex %2, %1, %3\n\t"
+			       "tst %2, %2\n\t"
+			       "bne 1b\n\t"
+			       : "=&r" ( old ), "=&r" ( new ), "=&l" ( flag ),
+				 "+Q" ( *dword )
+			       : "r" ( mask )
+			       : "cc" );
+
+	return ( old & mask );
+}
+
+/**
+ * Test and clear bit atomically
+ *
+ * @v bit		Bit to set
+ * @v bits		Bit field
+ * @ret old		Old value of bit (zero or non-zero)
+ */
+static inline __attribute__ (( always_inline )) int
+test_and_clear_bit ( unsigned int bit, volatile void *bits ) {
+	unsigned int index = ( bit / 32 );
+	unsigned int offset = ( bit % 32 );
+	volatile uint32_t *dword = ( ( ( volatile uint32_t * ) bits ) + index );
+	uint32_t mask = ( 1UL << offset );
+	uint32_t old;
+	uint32_t new;
+	uint32_t flag;
+
+	__asm__ __volatile__ ( "\n1:\n\t"
+			       "ldrex %0, %3\n\t"
+			       "bic %1, %0, %4\n\t"
+			       "strex %2, %1, %3\n\t"
+			       "tst %2, %2\n\t"
+			       "bne 1b\n\t"
+			       : "=&r" ( old ), "=&r" ( new ), "=&l" ( flag ),
+				 "+Q" ( *dword )
+			       : "r" ( mask )
+			       : "cc" );
+
+	return ( old & mask );
+}
+
+/**
+ * Set bit atomically
+ *
+ * @v bit		Bit to set
+ * @v bits		Bit field
+ */
+static inline __attribute__ (( always_inline )) void
+set_bit ( unsigned int bit, volatile void *bits ) {
+
+	test_and_set_bit ( bit, bits );
+}
+
+/**
+ * Clear bit atomically
+ *
+ * @v bit		Bit to set
+ * @v bits		Bit field
+ */
+static inline __attribute__ (( always_inline )) void
+clear_bit ( unsigned int bit, volatile void *bits ) {
+
+	test_and_clear_bit ( bit, bits );
+}
+
+#endif /* _BITS_BITOPS_H */

src/arch/arm32/include/bits/byteswap.h

@@ -0,0 +1,52 @@
+#ifndef _BITS_BYTESWAP_H
+#define _BITS_BYTESWAP_H
+
+/** @file
+ *
+ * Byte-order swapping functions
+ *
+ */
+
+#include <stdint.h>
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+static inline __attribute__ (( always_inline, const )) uint16_t
+__bswap_variable_16 ( uint16_t x ) {
+	__asm__ ( "rev16 %0, %1" : "=l" ( x ) : "l" ( x ) );
+	return x;
+}
+
+static inline __attribute__ (( always_inline )) void
+__bswap_16s ( uint16_t *x ) {
+	*x = __bswap_variable_16 ( *x );
+}
+
+static inline __attribute__ (( always_inline, const )) uint32_t
+__bswap_variable_32 ( uint32_t x ) {
+	__asm__ ( "rev %0, %1" : "=l" ( x ) : "l" ( x ) );
+	return x;
+}
+
+static inline __attribute__ (( always_inline )) void
+__bswap_32s ( uint32_t *x ) {
+	*x = __bswap_variable_32 ( *x );
+}
+
+static inline __attribute__ (( always_inline, const )) uint64_t
+__bswap_variable_64 ( uint64_t x ) {
+	uint32_t in_high = ( x >> 32 );
+	uint32_t in_low = ( x & 0xffffffffUL );
+	uint32_t out_high = __bswap_variable_32 ( in_low );
+	uint32_t out_low = __bswap_variable_32 ( in_high );
+
+	return ( ( ( ( uint64_t ) out_high ) << 32 ) |
+		 ( ( uint64_t ) out_low ) );
+}
+
+static inline __attribute__ (( always_inline )) void
+__bswap_64s ( uint64_t *x ) {
+	*x = __bswap_variable_64 ( *x );
+}
+
+#endif

src/arch/arm32/include/bits/compiler.h

@@ -0,0 +1,16 @@
+#ifndef _BITS_COMPILER_H
+#define _BITS_COMPILER_H
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/** Dummy relocation type */
+#define RELOC_TYPE_NONE R_ARM_NONE
+
+#ifndef ASSEMBLY
+
+#define __asmcall
+#define __libgcc
+
+#endif /* ASSEMBLY */
+
+#endif /*_BITS_COMPILER_H */

src/arch/arm32/include/bits/profile.h

@@ -0,0 +1,30 @@
+#ifndef _BITS_PROFILE_H
+#define _BITS_PROFILE_H
+
+/** @file
+ *
+ * Profiling
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+
+/**
+ * Get profiling timestamp
+ *
+ * @ret timestamp	Timestamp
+ */
+static inline __attribute__ (( always_inline )) uint64_t
+profile_timestamp ( void ) {
+	uint32_t cycles;
+
+	/* Read cycle counter */
+	__asm__ __volatile__ ( "mcr p15, 0, %1, c9, c12, 0\n\t"
+			       "mrc p15, 0, %0, c9, c13, 0\n\t"
+			       : "=r" ( cycles ) : "r" ( 1 ) );
+	return cycles;
+}
+
+#endif /* _BITS_PROFILE_H */

src/arch/arm32/include/bits/stdint.h

@@ -0,0 +1,23 @@
+#ifndef _BITS_STDINT_H
+#define _BITS_STDINT_H
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+typedef __SIZE_TYPE__		size_t;
+typedef signed long		ssize_t;
+typedef signed long		off_t;
+
+typedef unsigned char		uint8_t;
+typedef unsigned short		uint16_t;
+typedef unsigned int		uint32_t;
+typedef unsigned long long	uint64_t;
+
+typedef signed char		int8_t;
+typedef signed short		int16_t;
+typedef signed int		int32_t;
+typedef signed long long	int64_t;
+
+typedef unsigned long		physaddr_t;
+typedef unsigned long		intptr_t;
+
+#endif /* _BITS_STDINT_H */

src/arch/arm32/include/bits/string.h

@@ -0,0 +1,60 @@
+#ifndef BITS_STRING_H
+#define BITS_STRING_H
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/** @file
+ *
+ * String functions
+ *
+ */
+
+/**
+ * Fill memory region
+ *
+ * @v dest		Destination region
+ * @v character		Fill character
+ * @v len		Length
+ * @ret dest		Destination region
+ */
+static inline __attribute__ (( always_inline )) void *
+memset ( void *dest, int character, size_t len ) {
+
+	/* Not yet optimised */
+	generic_memset ( dest, character, len );
+	return dest;
+}
+
+/**
+ * Copy memory region
+ *
+ * @v dest		Destination region
+ * @v src		Source region
+ * @v len		Length
+ * @ret dest		Destination region
+ */
+static inline __attribute__ (( always_inline )) void *
+memcpy ( void *dest, const void *src, size_t len ) {
+
+	/* Not yet optimised */
+	generic_memcpy ( dest, src, len );
+	return dest;
+}
+
+/**
+ * Copy (possibly overlapping) memory region
+ *
+ * @v dest		Destination region
+ * @v src		Source region
+ * @v len		Length
+ * @ret dest		Destination region
+ */
+static inline __attribute__ (( always_inline )) void *
+memmove ( void *dest, const void *src, size_t len ) {
+
+	/* Not yet optimised */
+	generic_memmove ( dest, src, len );
+	return dest;
+}
+
+#endif /* BITS_STRING_H */

src/arch/arm32/include/bits/strings.h

@@ -0,0 +1,85 @@
+#ifndef _BITS_STRINGS_H
+#define _BITS_STRINGS_H
+
+/** @file
+ *
+ * String functions
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/**
+ * Find first (i.e. least significant) set bit
+ *
+ * @v value		Value
+ * @ret lsb		Least significant bit set in value (LSB=1), or zero
+ */
+static inline __attribute__ (( always_inline )) int __ffsl ( long value ) {
+	unsigned long bits = value;
+	unsigned long lsb;
+	unsigned int lz;
+
+	/* Extract least significant set bit */
+	lsb = ( bits & -bits );
+
+	/* Count number of leading zeroes before LSB */
+	__asm__ ( "clz %0, %1" : "=r" ( lz ) : "r" ( lsb ) );
+
+	return ( 32 - lz );
+}
+
+/**
+ * Find first (i.e. least significant) set bit
+ *
+ * @v value		Value
+ * @ret lsb		Least significant bit set in value (LSB=1), or zero
+ */
+static inline __attribute__ (( always_inline )) int __ffsll ( long long value ){
+	unsigned long high = ( value >> 32 );
+	unsigned long low = ( value >> 0 );
+
+	if ( low ) {
+		return ( __ffsl ( low ) );
+	} else if ( high ) {
+		return ( 32 + __ffsl ( high ) );
+	} else {
+		return 0;
+	}
+}
+
+/**
+ * Find last (i.e. most significant) set bit
+ *
+ * @v value		Value
+ * @ret msb		Most significant bit set in value (LSB=1), or zero
+ */
+static inline __attribute__ (( always_inline )) int __flsl ( long value ) {
+	unsigned int lz;
+
+	/* Count number of leading zeroes */
+	__asm__ ( "clz %0, %1" : "=r" ( lz ) : "r" ( value ) );
+
+	return ( 32 - lz );
+}
+
+/**
+ * Find last (i.e. most significant) set bit
+ *
+ * @v value		Value
+ * @ret msb		Most significant bit set in value (LSB=1), or zero
+ */
+static inline __attribute__ (( always_inline )) int __flsll ( long long value ){
+	unsigned long high = ( value >> 32 );
+	unsigned long low = ( value >> 0 );
+
+	if ( high ) {
+		return ( 32 + __flsl ( high ) );
+	} else if ( low ) {
+		return ( __flsl ( low ) );
+	} else {
+		return 0;
+	}
+}
+
+#endif /* _BITS_STRINGS_H */

src/arch/arm32/include/bits/tcpip.h

@@ -0,0 +1,19 @@
+#ifndef _BITS_TCPIP_H
+#define _BITS_TCPIP_H
+
+/** @file
+ *
+ * Transport-network layer interface
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+static inline __attribute__ (( always_inline )) uint16_t
+tcpip_continue_chksum ( uint16_t partial, const void *data, size_t len ) {
+
+	/* Not yet optimised */
+	return generic_tcpip_continue_chksum ( partial, data, len );
+}
+
+#endif /* _BITS_TCPIP_H */

src/arch/arm32/include/efi/ipxe/dhcp_arch.h

@@ -0,0 +1,40 @@
+/*
+ * Copyright (C) 2015 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+#ifndef _DHCP_ARCH_H
+#define _DHCP_ARCH_H
+
+/** @file
+ *
+ * Architecture-specific DHCP options
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <ipxe/dhcp.h>
+
+#define DHCP_ARCH_CLIENT_ARCHITECTURE DHCP_CLIENT_ARCHITECTURE_ARM32
+
+#define DHCP_ARCH_CLIENT_NDI 1 /* UNDI */ , 3, 10 /* v3.10 */
+
+#endif

src/arch/arm32/include/gdbmach.h

@@ -0,0 +1,45 @@
+#ifndef GDBMACH_H
+#define GDBMACH_H
+
+/** @file
+ *
+ * GDB architecture specifics
+ *
+ * This file declares functions for manipulating the machine state and
+ * debugging context.
+ *
+ */
+
+#include <stdint.h>
+
+typedef unsigned long gdbreg_t;
+
+/* Register snapshot */
+enum {
+	/* Not yet implemented */
+	GDBMACH_NREGS,
+};
+
+#define GDBMACH_SIZEOF_REGS ( GDBMACH_NREGS * sizeof ( gdbreg_t ) )
+
+static inline void gdbmach_set_pc ( gdbreg_t *regs, gdbreg_t pc ) {
+	/* Not yet implemented */
+	( void ) regs;
+	( void ) pc;
+}
+
+static inline void gdbmach_set_single_step ( gdbreg_t *regs, int step ) {
+	/* Not yet implemented */
+	( void ) regs;
+	( void ) step;
+}
+
+static inline void gdbmach_breakpoint ( void ) {
+	/* Not yet implemented */
+}
+
+extern int gdbmach_set_breakpoint ( int type, unsigned long addr, size_t len,
+				    int enable );
+extern void gdbmach_init ( void );
+
+#endif /* GDBMACH_H */

src/arch/arm32/include/limits.h

@@ -0,0 +1,61 @@
+#ifndef LIMITS_H
+#define LIMITS_H	1
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/* Number of bits in a `char' */
+#define CHAR_BIT	8
+
+/* Minimum and maximum values a `signed char' can hold */
+#define SCHAR_MIN	(-128)
+#define SCHAR_MAX	127
+
+/* Maximum value an `unsigned char' can hold. (Minimum is 0.) */
+#define UCHAR_MAX	255
+
+/* Minimum and maximum values a `char' can hold */
+#define CHAR_MIN	SCHAR_MIN
+#define CHAR_MAX	SCHAR_MAX
+
+/* Minimum and maximum values a `signed short int' can hold */
+#define SHRT_MIN	(-32768)
+#define SHRT_MAX	32767
+
+/* Maximum value an `unsigned short' can hold. (Minimum is 0.) */
+#define USHRT_MAX	65535
+
+
+/* Minimum and maximum values a `signed int' can hold */
+#define INT_MIN		(-INT_MAX - 1)
+#define INT_MAX		2147483647
+
+/* Maximum value an `unsigned int' can hold. (Minimum is 0.) */
+#define UINT_MAX	4294967295U
+
+
+/* Minimum and maximum values a `signed int' can hold */
+#define INT_MAX		2147483647
+#define INT_MIN		(-INT_MAX - 1)
+
+
+/* Maximum value an `unsigned int' can hold. (Minimum is 0.) */
+#define UINT_MAX	4294967295U
+
+
+/* Minimum and maximum values a `signed long' can hold */
+#define LONG_MAX	2147483647
+#define LONG_MIN	(-LONG_MAX - 1L)
+
+/* Maximum value an `unsigned long' can hold. (Minimum is 0.) */
+#define ULONG_MAX	4294967295UL
+
+/* Minimum and maximum values a `signed long long' can hold */
+#define LLONG_MAX	9223372036854775807LL
+#define LLONG_MIN	(-LONG_MAX - 1LL)
+
+
+/* Maximum value an `unsigned long long' can hold. (Minimum is 0.) */
+#define ULLONG_MAX	18446744073709551615ULL
+
+
+#endif /* LIMITS_H */

src/arch/arm32/include/setjmp.h

@@ -0,0 +1,38 @@
+#ifndef _SETJMP_H
+#define _SETJMP_H
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+
+/** A jump buffer */
+typedef struct {
+	/** Saved r4 */
+	uint32_t r4;
+	/** Saved r5 */
+	uint32_t r5;
+	/** Saved r6 */
+	uint32_t r6;
+	/** Saved r7 */
+	uint32_t r7;
+	/** Saved r8 */
+	uint32_t r8;
+	/** Saved r9 */
+	uint32_t r9;
+	/** Saved r10 */
+	uint32_t r10;
+	/** Saved frame pointer (r11) */
+	uint32_t fp;
+	/** Saved stack pointer (r13) */
+	uint32_t sp;
+	/** Saved link register (r14) */
+	uint32_t lr;
+} jmp_buf[1];
+
+extern int __asmcall __attribute__ (( returns_twice ))
+setjmp ( jmp_buf env );
+
+extern void __asmcall __attribute__ (( noreturn ))
+longjmp ( jmp_buf env, int val );
+
+#endif /* _SETJMP_H */

src/arch/arm32/libgcc/lldivmod.S

@@ -0,0 +1,50 @@
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL )
+
+	.text
+	.thumb
+
+/**
+ * Unsigned long long division
+ *
+ * @v r1:r0		Dividend
+ * @v r3:r2		Divisor
+ * @ret r1:r0		Quotient
+ * @ret r3:r2		Remainder
+ */
+	.section ".text.__aeabi_uldivmod", "ax", %progbits
+	.globl	__aeabi_uldivmod
+	.type	__aeabi_uldivmod, %function
+__aeabi_uldivmod:
+	/* Allocate stack space for remainder and pointer to remainder */
+	push	{r0, r1, r2, r3, r4, lr}
+	/* Call __udivmoddi4() */
+	add	r4, sp, #8
+	str	r4, [sp]
+	bl	__udivmoddi4
+	/* Retrieve remainder and return */
+	add	sp, sp, #8
+	pop	{r2, r3, r4, pc}
+	.size	__aeabi_uldivmod, . - __aeabi_uldivmod
+
+/**
+ * Signed long long division
+ *
+ * @v r1:r0		Dividend
+ * @v r3:r2		Divisor
+ * @ret r1:r0		Quotient
+ * @ret r3:r2		Remainder
+ */
+	.section ".text.__aeabi_ldivmod", "ax", %progbits
+	.globl	__aeabi_ldivmod
+	.type	__aeabi_ldivmod, %function
+__aeabi_ldivmod:
+	/* Allocate stack space for remainder and pointer to remainder */
+	push	{r0, r1, r2, r3, r4, lr}
+	/* Call __divmoddi4() */
+	add	r4, sp, #8
+	str	r4, [sp]
+	bl	__divmoddi4
+	/* Retrieve remainder and return */
+	add	sp, sp, #8
+	pop	{r2, r3, r4, pc}
+	.size	__aeabi_ldivmod, . - __aeabi_ldivmod

src/arch/arm32/libgcc/llshift.S

@@ -0,0 +1,88 @@
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL )
+
+	.text
+	.arm
+
+/**
+ * Logical shift left
+ *
+ * @v r1:r0		Value to shift
+ * @v r2		Shift amount
+ * @ret r1:r0		Shifted value
+ */
+	.section ".text.__aeabi_llsl", "ax", %progbits
+	.globl	__aeabi_llsl
+	.type	__aeabi_llsl, %function
+__aeabi_llsl:
+	/* r3 = ( shift - 32 ) */
+	subs	r3, r2, #32
+	/* If shift >= 32, then
+	 *   high = ( low << ( shift - 32 ) )
+	 */
+	movpl	r1, r0, lsl r3
+	/* If shift < 32, then
+	 *   high = ( ( high << shift ) | ( low >> ( 32 - shift ) ) )
+	 */
+	movmi	r1, r1, lsl r2
+	rsbmi	r3, r2, #32
+	orrmi	r1, r1, r0, lsr r3
+	/* low = ( low << shift ) */
+	mov	r0, r0, lsl r2
+	bx	lr
+	.size	__aeabi_llsl, . - __aeabi_llsl
+
+/**
+ * Logical shift right
+ *
+ * @v r1:r0		Value to shift
+ * @v r2		Shift amount
+ * @ret r1:r0		Shifted value
+ */
+	.section ".text.__aeabi_llsr", "ax", %progbits
+	.globl	__aeabi_llsr
+	.type	__aeabi_llsr, %function
+__aeabi_llsr:
+	/* r3 = ( shift - 32 ) */
+	subs	r3, r2, #32
+	/* If shift >= 32, then
+	 *   low = ( high >> ( shift - 32 ) )
+	 */
+	movpl	r0, r1, lsr r3
+	/* If shift < 32, then
+	 *   low = ( ( low >> shift ) | ( high << ( 32 - shift ) ) )
+	 */
+	movmi	r0, r0, lsr r2
+	rsbmi	r3, r2, #32
+	orrmi	r0, r0, r1, lsl r3
+	/* high = ( high >> shift ) */
+	mov	r1, r1, lsr r2
+	bx	lr
+	.size	__aeabi_llsr, . - __aeabi_llsr
+
+/**
+ * Arithmetic shift right
+ *
+ * @v r1:r0		Value to shift
+ * @v r2		Shift amount
+ * @ret r1:r0		Shifted value
+ */
+	.section ".text.__aeabi_lasr", "ax", %progbits
+	.globl	__aeabi_lasr
+	.type	__aeabi_lasr, %function
+__aeabi_lasr:
+	/* r3 = ( shift - 32 ) */
+	subs	r3, r2, #32
+	/* If shift >= 32, then
+	 *   low = ( high >> ( shift - 32 ) )
+	 */
+	movpl	r0, r1, asr r3
+	/* If shift < 32, then
+	 *   low = ( ( low >> shift ) | ( high << ( 32 - shift ) ) )
+	 */
+	movmi	r0, r0, lsr r2
+	rsbmi	r3, r2, #32
+	orrmi	r0, r0, r1, lsl r3
+	/* high = ( high >> shift ) */
+	mov	r1, r1, asr r2
+	bx	lr
+	.size	__aeabi_lasr, . - __aeabi_lasr

src/arch/arm64/Makefile

@@ -0,0 +1,33 @@
+# ARM64-specific directories containing source files
+#
+SRCDIRS		+= arch/arm64/core
+
+# ARM64-specific flags
+#
+CFLAGS		+= -mlittle-endian -mcmodel=small
+CFLAGS		+= -fomit-frame-pointer
+ASFLAGS		+= -mabi=lp64 -EL
+
+# We want to specify the LP64 model.  There is an explicit -mabi=lp64
+# on GCC 4.9 and later, and no guarantee as to which is the default
+# model.  In earlier versions of GCC, there is no -mabi option and the
+# default appears to be LP64 anyway.
+#
+ifeq ($(CCTYPE),gcc)
+LP64_TEST = $(CC) -mabi=lp64 -x c -c /dev/null -o /dev/null >/dev/null 2>&1
+LP64_FLAGS := $(shell $(LP64_TEST) && $(ECHO) '-mabi=lp64')
+WORKAROUND_CFLAGS += $(LP64_FLAGS)
+endif
+
+# EFI requires -fshort-wchar, and nothing else currently uses wchar_t
+#
+CFLAGS		+= -fshort-wchar
+
+# Include common ARM Makefile
+MAKEDEPS	+= arch/arm/Makefile
+include arch/arm/Makefile
+
+# Include platform-specific Makefile
+#
+MAKEDEPS	+= arch/arm64/Makefile.$(PLATFORM)
+include arch/arm64/Makefile.$(PLATFORM)

src/arch/arm64/Makefile.efi

@@ -0,0 +1,18 @@
+# -*- makefile -*- : Force emacs to use Makefile mode
+
+# Avoid untranslatable relocations
+#
+CFLAGS		+= -fno-pic
+
+# Specify EFI image builder
+#
+ELF2EFI		= $(ELF2EFI64)
+
+# Specify EFI boot file
+#
+EFI_BOOT_FILE	= bootaa64.efi
+
+# Include generic EFI Makefile
+#
+MAKEDEPS	+= arch/arm/Makefile.efi
+include arch/arm/Makefile.efi

src/arch/arm64/core/arm64_bigint.c

@@ -0,0 +1,103 @@
+/*
+ * Copyright (C) 2016 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+#include <string.h>
+#include <ipxe/bigint.h>
+
+/** @file
+ *
+ * Big integer support
+ */
+
+/**
+ * Multiply big integers
+ *
+ * @v multiplicand0	Element 0 of big integer to be multiplied
+ * @v multiplier0	Element 0 of big integer to be multiplied
+ * @v result0		Element 0 of big integer to hold result
+ * @v size		Number of elements
+ */
+void bigint_multiply_raw ( const uint64_t *multiplicand0,
+			   const uint64_t *multiplier0,
+			   uint64_t *result0, unsigned int size ) {
+	const bigint_t ( size ) __attribute__ (( may_alias )) *multiplicand =
+		( ( const void * ) multiplicand0 );
+	const bigint_t ( size ) __attribute__ (( may_alias )) *multiplier =
+		( ( const void * ) multiplier0 );
+	bigint_t ( size * 2 ) __attribute__ (( may_alias )) *result =
+		( ( void * ) result0 );
+	unsigned int i;
+	unsigned int j;
+	uint64_t multiplicand_element;
+	uint64_t multiplier_element;
+	uint64_t *result_elements;
+	uint64_t discard_low;
+	uint64_t discard_high;
+	uint64_t discard_temp_low;
+	uint64_t discard_temp_high;
+
+	/* Zero result */
+	memset ( result, 0, sizeof ( *result ) );
+
+	/* Multiply integers one element at a time */
+	for ( i = 0 ; i < size ; i++ ) {
+		multiplicand_element = multiplicand->element[i];
+		for ( j = 0 ; j < size ; j++ ) {
+			multiplier_element = multiplier->element[j];
+			result_elements = &result->element[ i + j ];
+			/* Perform a single multiply, and add the
+			 * resulting double-element into the result,
+			 * carrying as necessary.  The carry can
+			 * never overflow beyond the end of the
+			 * result, since:
+			 *
+			 *     a < 2^{n}, b < 2^{n} => ab < 2^{2n}
+			 */
+			__asm__ __volatile__ ( "mul %1, %6, %7\n\t"
+					       "umulh %2, %6, %7\n\t"
+					       "ldp %3, %4, [%0]\n\t"
+					       "adds %3, %3, %1\n\t"
+					       "adcs %4, %4, %2\n\t"
+					       "stp %3, %4, [%0], #16\n\t"
+					       "bcc 2f\n\t"
+					       "\n1:\n\t"
+					       "ldr %3, [%0]\n\t"
+					       "adcs %3, %3, xzr\n\t"
+					       "str %3, [%0], #8\n\t"
+					       "bcs 1b\n\t"
+					       "\n2:\n\t"
+					       : "+r" ( result_elements ),
+						 "=&r" ( discard_low ),
+						 "=&r" ( discard_high ),
+						 "=r" ( discard_temp_low ),
+						 "=r" ( discard_temp_high ),
+						 "+m" ( *result )
+					       : "r" ( multiplicand_element ),
+						 "r" ( multiplier_element )
+					       : "cc" );
+		}
+	}
+}

src/arch/arm64/core/arm64_string.c

@@ -0,0 +1,249 @@
+/*
+ * Copyright (C) 2016 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+/** @file
+ *
+ * Optimised string operations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <string.h>
+
+/**
+ * Copy memory area
+ *
+ * @v dest		Destination address
+ * @v src		Source address
+ * @v len		Length
+ * @ret dest		Destination address
+ */
+void arm64_memcpy ( void *dest, const void *src, size_t len ) {
+	void *discard_dest;
+	void *discard_end;
+	const void *discard_src;
+	size_t discard_offset;
+	unsigned long discard_data;
+	unsigned long discard_low;
+	unsigned long discard_high;
+
+	/* If length is too short for an "ldp"/"stp" instruction pair,
+	 * then just copy individual bytes.
+	 */
+	if ( len < 16 ) {
+		__asm__ __volatile__ ( "cbz %0, 2f\n\t"
+				       "\n1:\n\t"
+				       "sub %0, %0, #1\n\t"
+				       "ldrb %w1, [%3, %0]\n\t"
+				       "strb %w1, [%2, %0]\n\t"
+				       "cbnz %0, 1b\n\t"
+				       "\n2:\n\t"
+				       : "=&r" ( discard_offset ),
+					 "=&r" ( discard_data )
+				       : "r" ( dest ), "r" ( src ), "0" ( len )
+				       : "memory" );
+		return;
+	}
+
+	/* Use "ldp"/"stp" to copy 16 bytes at a time: one initial
+	 * potentially unaligned access, multiple destination-aligned
+	 * accesses, one final potentially unaligned access.
+	 */
+	__asm__ __volatile__ ( "ldp %3, %4, [%1], #16\n\t"
+			       "stp %3, %4, [%0], #16\n\t"
+			       "and %3, %0, #15\n\t"
+			       "sub %0, %0, %3\n\t"
+			       "sub %1, %1, %3\n\t"
+			       "bic %2, %5, #15\n\t"
+			       "b 2f\n\t"
+			       "\n1:\n\t"
+			       "ldp %3, %4, [%1], #16\n\t"
+			       "stp %3, %4, [%0], #16\n\t"
+			       "\n2:\n\t"
+			       "cmp %0, %2\n\t"
+			       "bne 1b\n\t"
+			       "ldp %3, %4, [%6, #-16]\n\t"
+			       "stp %3, %4, [%5, #-16]\n\t"
+			       : "=&r" ( discard_dest ),
+				 "=&r" ( discard_src ),
+				 "=&r" ( discard_end ),
+				 "=&r" ( discard_low ),
+				 "=&r" ( discard_high )
+			       : "r" ( dest + len ), "r" ( src + len ),
+				 "0" ( dest ), "1" ( src )
+			       : "memory", "cc" );
+}
+
+/**
+ * Zero memory region
+ *
+ * @v dest		Destination region
+ * @v len		Length
+ */
+void arm64_bzero ( void *dest, size_t len ) {
+	size_t discard_offset;
+	void *discard_dest;
+	void *discard_end;
+
+	/* If length is too short for an "stp" instruction, then just
+	 * zero individual bytes.
+	 */
+	if ( len < 16 ) {
+		__asm__ __volatile__ ( "cbz %0, 2f\n\t"
+				       "\n1:\n\t"
+				       "sub %0, %0, #1\n\t"
+				       "strb wzr, [%1, %0]\n\t"
+				       "cbnz %0, 1b\n\t"
+				       "\n2:\n\t"
+				       : "=&r" ( discard_offset )
+				       : "r" ( dest ), "0" ( len )
+				       : "memory" );
+		return;
+	}
+
+	/* Use "stp" to zero 16 bytes at a time: one initial
+	 * potentially unaligned access, multiple aligned accesses,
+	 * one final potentially unaligned access.
+	 */
+	__asm__ __volatile__ ( "stp xzr, xzr, [%0], #16\n\t"
+			       "bic %0, %0, #15\n\t"
+			       "bic %1, %2, #15\n\t"
+			       "b 2f\n\t"
+			       "\n1:\n\t"
+			       "stp xzr, xzr, [%0], #16\n\t"
+			       "\n2:\n\t"
+			       "cmp %0, %1\n\t"
+			       "bne 1b\n\t"
+			       "stp xzr, xzr, [%2, #-16]\n\t"
+			       : "=&r" ( discard_dest ),
+				 "=&r" ( discard_end )
+			       : "r" ( dest + len ), "0" ( dest )
+			       : "memory", "cc" );
+}
+
+/**
+ * Fill memory region
+ *
+ * @v dest		Destination region
+ * @v len		Length
+ * @v character		Fill character
+ *
+ * The unusual parameter order is to allow for more efficient
+ * tail-calling to arm64_memset() when zeroing a region.
+ */
+void arm64_memset ( void *dest, size_t len, int character ) {
+	size_t discard_offset;
+
+	/* Use optimised zeroing code if applicable */
+	if ( character == 0 ) {
+		arm64_bzero ( dest, len );
+		return;
+	}
+
+	/* Fill one byte at a time.  Calling memset() with a non-zero
+	 * value is relatively rare and unlikely to be
+	 * performance-critical.
+	 */
+	__asm__ __volatile__ ( "cbz %0, 2f\n\t"
+			       "\n1:\n\t"
+			       "sub %0, %0, #1\n\t"
+			       "strb %w2, [%1, %0]\n\t"
+			       "cbnz %0, 1b\n\t"
+			       "\n2:\n\t"
+			       : "=&r" ( discard_offset )
+			       : "r" ( dest ), "r" ( character ), "0" ( len )
+			       : "memory" );
+}
+
+/**
+ * Copy (possibly overlapping) memory region forwards
+ *
+ * @v dest		Destination region
+ * @v src		Source region
+ * @v len		Length
+ */
+void arm64_memmove_forwards ( void *dest, const void *src, size_t len ) {
+	void *discard_dest;
+	const void *discard_src;
+	unsigned long discard_data;
+
+	/* Assume memmove() is not performance-critical, and perform a
+	 * bytewise copy for simplicity.
+	 */
+	__asm__ __volatile__ ( "b 2f\n\t"
+			       "\n1:\n\t"
+			       "ldrb %w2, [%1], #1\n\t"
+			       "strb %w2, [%0], #1\n\t"
+			       "\n2:\n\t"
+			       "cmp %0, %3\n\t"
+			       "bne 1b\n\t"
+			       : "=&r" ( discard_dest ),
+				 "=&r" ( discard_src ),
+				 "=&r" ( discard_data )
+			       : "r" ( dest + len ), "0" ( dest ), "1" ( src )
+			       : "memory" );
+}
+
+/**
+ * Copy (possibly overlapping) memory region backwards
+ *
+ * @v dest		Destination region
+ * @v src		Source region
+ * @v len		Length
+ */
+void arm64_memmove_backwards ( void *dest, const void *src, size_t len ) {
+	size_t discard_offset;
+	unsigned long discard_data;
+
+	/* Assume memmove() is not performance-critical, and perform a
+	 * bytewise copy for simplicity.
+	 */
+	__asm__ __volatile__ ( "cbz %0, 2f\n\t"
+			       "\n1:\n\t"
+			       "sub %0, %0, #1\n\t"
+			       "ldrb %w1, [%3, %0]\n\t"
+			       "strb %w1, [%2, %0]\n\t"
+			       "cbnz %0, 1b\n\t"
+			       "\n2:\n\t"
+			       : "=&r" ( discard_offset ),
+				 "=&r" ( discard_data )
+			       : "r" ( dest ), "r" ( src ), "0" ( len )
+			       : "memory" );
+}
+
+/**
+ * Copy (possibly overlapping) memory region
+ *
+ * @v dest		Destination region
+ * @v src		Source region
+ * @v len		Length
+ */
+void arm64_memmove ( void *dest, const void *src, size_t len ) {
+
+	if ( dest <= src ) {
+		arm64_memmove_forwards ( dest, src, len );
+	} else {
+		arm64_memmove_backwards ( dest, src, len );
+	}
+}

src/arch/arm64/core/arm64_tcpip.c

@@ -0,0 +1,175 @@
+/*
+ * Copyright (C) 2016 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/** @file
+ *
+ * TCP/IP checksum
+ *
+ */
+
+#include <strings.h>
+#include <ipxe/tcpip.h>
+
+/** Alignment used by main checksumming loop */
+#define TCPIP_CHKSUM_ALIGN 16
+
+/** Number of steps in each iteration of the unrolled main checksumming loop */
+#define TCPIP_CHKSUM_UNROLL 4
+
+/**
+ * Calculate continued TCP/IP checkum
+ *
+ * @v sum		Checksum of already-summed data, in network byte order
+ * @v data		Data buffer
+ * @v len		Length of data buffer
+ * @ret sum		Updated checksum, in network byte order
+ */
+uint16_t tcpip_continue_chksum ( uint16_t sum, const void *data,
+				 size_t len ) {
+	intptr_t start;
+	intptr_t end;
+	intptr_t mid;
+	unsigned int pre;
+	unsigned int post;
+	unsigned int first;
+	uint64_t discard_low;
+	uint64_t discard_high;
+
+	/* Avoid potentially undefined shift operation */
+	if ( len == 0 )
+		return sum;
+
+	/* Find maximally-aligned midpoint.  For short blocks of data,
+	 * this may be aligned to fewer than 16 bytes.
+	 */
+	start = ( ( intptr_t ) data );
+	end = ( start + len );
+	mid = ( end &
+		~( ( ~( 1UL << 63 ) ) >> ( 64 - flsl ( start ^ end ) ) ) );
+
+	/* Calculate pre- and post-alignment lengths */
+	pre = ( ( mid - start ) & ( TCPIP_CHKSUM_ALIGN - 1 ) );
+	post = ( ( end - mid ) & ( TCPIP_CHKSUM_ALIGN - 1 ) );
+
+	/* Calculate number of steps in first iteration of unrolled loop */
+	first = ( ( ( len - pre - post ) / TCPIP_CHKSUM_ALIGN ) &
+		  ( TCPIP_CHKSUM_UNROLL - 1 ) );
+
+	/* Calculate checksum */
+	__asm__ ( /* Invert sum */
+		  "eor %w0, %w0, #0xffff\n\t"
+		  /* Clear carry flag */
+		  "cmn xzr, xzr\n\t"
+		  /* Byteswap and sum pre-alignment byte, if applicable */
+		  "tbz %w4, #0, 1f\n\t"
+		  "ldrb %w2, [%1], #1\n\t"
+		  "rev16 %w0, %w0\n\t"
+		  "rev16 %w2, %w2\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "\n1:\n\t"
+		  /* Sum pre-alignment halfword, if applicable */
+		  "tbz %w4, #1, 1f\n\t"
+		  "ldrh %w2, [%1], #2\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "\n1:\n\t"
+		  /* Sum pre-alignment word, if applicable */
+		  "tbz %w4, #2, 1f\n\t"
+		  "ldr %w2, [%1], #4\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "\n1:\n\t"
+		  /* Sum pre-alignment doubleword, if applicable */
+		  "tbz %w4, #3, 1f\n\t"
+		  "ldr %2, [%1], #8\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "\n1:\n\t"
+		  /* Jump into unrolled (x4) main loop */
+		  "adr %2, 2f\n\t"
+		  "sub %2, %2, %5, lsl #3\n\t"
+		  "sub %2, %2, %5, lsl #2\n\t"
+		  "br %2\n\t"
+		  "\n1:\n\t"
+		  "ldp %2, %3, [%1], #16\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "adcs %0, %0, %3\n\t"
+		  "ldp %2, %3, [%1], #16\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "adcs %0, %0, %3\n\t"
+		  "ldp %2, %3, [%1], #16\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "adcs %0, %0, %3\n\t"
+		  "ldp %2, %3, [%1], #16\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "adcs %0, %0, %3\n\t"
+		  "\n2:\n\t"
+		  "sub %2, %1, %6\n\t"
+		  "cbnz %2, 1b\n\t"
+		  /* Sum post-alignment doubleword, if applicable */
+		  "tbz %w7, #3, 1f\n\t"
+		  "ldr %2, [%1], #8\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "\n1:\n\t"
+		  /* Sum post-alignment word, if applicable */
+		  "tbz %w7, #2, 1f\n\t"
+		  "ldr %w2, [%1], #4\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "\n1:\n\t"
+		  /* Sum post-alignment halfword, if applicable */
+		  "tbz %w7, #1, 1f\n\t"
+		  "ldrh %w2, [%1], #2\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "\n1:\n\t"
+		  /* Sum post-alignment byte, if applicable */
+		  "tbz %w7, #0, 1f\n\t"
+		  "ldrb %w2, [%1], #1\n\t"
+		  "adcs %0, %0, %2\n\t"
+		  "\n1:\n\t"
+		  /* Fold down to a uint32_t plus carry flag */
+		  "lsr %2, %0, #32\n\t"
+		  "adcs %w0, %w0, %w2\n\t"
+		  /* Fold down to a uint16_t plus carry in bit 16 */
+		  "ubfm %2, %0, #0, #15\n\t"
+		  "ubfm %3, %0, #16, #31\n\t"
+		  "adc %w0, %w2, %w3\n\t"
+		  /* Fold down to a uint16_t */
+		  "tbz %w0, #16, 1f\n\t"
+		  "mov %w2, #0xffff\n\t"
+		  "sub %w0, %w0, %w2\n\t"
+		  "tbz %w0, #16, 1f\n\t"
+		  "sub %w0, %w0, %w2\n\t"
+		  "\n1:\n\t"
+		  /* Byteswap back, if applicable */
+		  "tbz %w4, #0, 1f\n\t"
+		  "rev16 %w0, %w0\n\t"
+		  "\n1:\n\t"
+		  /* Invert sum */
+		  "eor %w0, %w0, #0xffff\n\t"
+		  : "+r" ( sum ), "+r" ( data ), "=&r" ( discard_low ),
+		    "=&r" ( discard_high )
+		  : "r" ( pre ), "r" ( first ), "r" ( end - post ),
+		    "r" ( post )
+		  : "cc" );
+
+	return sum;
+}

src/arch/arm64/core/setjmp.S

@@ -0,0 +1,56 @@
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL )
+
+	.text
+
+	/* Must match jmp_buf structure layout */
+	.struct	0
+env_x19_x20:	.quad	0, 0
+env_x21_x22:	.quad	0, 0
+env_x23_x24:	.quad	0, 0
+env_x25_x26:	.quad	0, 0
+env_x27_x28:	.quad	0, 0
+env_x29_x30:	.quad	0, 0
+env_sp:		.quad	0
+	.previous
+
+/*
+ * Save stack context for non-local goto
+ */
+	.globl	setjmp
+	.type	setjmp, %function
+setjmp:
+	/* Store registers */
+	stp	x19, x20, [x0, #env_x19_x20]
+	stp	x21, x22, [x0, #env_x21_x22]
+	stp	x23, x24, [x0, #env_x23_x24]
+	stp	x25, x26, [x0, #env_x25_x26]
+	stp	x27, x28, [x0, #env_x27_x28]
+	stp	x29, x30, [x0, #env_x29_x30]
+	mov	x16, sp
+	str	x16, [x0, #env_sp]
+	/* Return 0 when returning as setjmp() */
+	mov	x0, #0
+	ret
+	.size	setjmp, . - setjmp
+
+/*
+ * Non-local jump to a saved stack context
+ */
+	.globl	longjmp
+	.type	longjmp, %function
+longjmp:
+	/* Restore registers */
+	ldp	x19, x20, [x0, #env_x19_x20]
+	ldp	x21, x22, [x0, #env_x21_x22]
+	ldp	x23, x24, [x0, #env_x23_x24]
+	ldp	x25, x26, [x0, #env_x25_x26]
+	ldp	x27, x28, [x0, #env_x27_x28]
+	ldp	x29, x30, [x0, #env_x29_x30]
+	ldr	x16, [x0, #env_sp]
+	mov	sp, x16
+	/* Force result to non-zero */
+	cmp	w1, #0
+	csinc	w0, w1, w1, ne
+	/* Return to setjmp() caller */
+	br	x30
+	.size	longjmp, . - longjmp

src/arch/arm64/include/bits/bigint.h

@@ -0,0 +1,317 @@
+#ifndef _BITS_BIGINT_H
+#define _BITS_BIGINT_H
+
+/** @file
+ *
+ * Big integer support
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+#include <string.h>
+#include <strings.h>
+
+/** Element of a big integer */
+typedef uint64_t bigint_element_t;
+
+/**
+ * Initialise big integer
+ *
+ * @v value0		Element 0 of big integer to initialise
+ * @v size		Number of elements
+ * @v data		Raw data
+ * @v len		Length of raw data
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_init_raw ( uint64_t *value0, unsigned int size,
+		  const void *data, size_t len ) {
+	size_t pad_len = ( sizeof ( bigint_t ( size ) ) - len );
+	uint8_t *value_byte = ( ( void * ) value0 );
+	const uint8_t *data_byte = ( data + len );
+
+	/* Copy raw data in reverse order, padding with zeros */
+	while ( len-- )
+		*(value_byte++) = *(--data_byte);
+	while ( pad_len-- )
+		*(value_byte++) = 0;
+}
+
+/**
+ * Add big integers
+ *
+ * @v addend0		Element 0 of big integer to add
+ * @v value0		Element 0 of big integer to be added to
+ * @v size		Number of elements
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_add_raw ( const uint64_t *addend0, uint64_t *value0,
+		 unsigned int size ) {
+	bigint_t ( size ) __attribute__ (( may_alias )) *value =
+		( ( void * ) value0 );
+	uint64_t *discard_addend;
+	uint64_t *discard_value;
+	uint64_t discard_addend_i;
+	uint64_t discard_value_i;
+	unsigned int discard_size;
+
+	__asm__ __volatile__ ( "cmn xzr, xzr\n\t" /* clear CF */
+			       "\n1:\n\t"
+			       "ldr %3, [%0], #8\n\t"
+			       "ldr %4, [%1]\n\t"
+			       "adcs %4, %4, %3\n\t"
+			       "str %4, [%1], #8\n\t"
+			       "sub %w2, %w2, #1\n\t"
+			       "cbnz %w2, 1b\n\t"
+			       : "=r" ( discard_addend ),
+				 "=r" ( discard_value ),
+				 "=r" ( discard_size ),
+				 "=r" ( discard_addend_i ),
+				 "=r" ( discard_value_i ),
+				 "+m" ( *value )
+			       : "0" ( addend0 ), "1" ( value0 ), "2" ( size )
+			       : "cc" );
+}
+
+/**
+ * Subtract big integers
+ *
+ * @v subtrahend0	Element 0 of big integer to subtract
+ * @v value0		Element 0 of big integer to be subtracted from
+ * @v size		Number of elements
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_subtract_raw ( const uint64_t *subtrahend0, uint64_t *value0,
+		      unsigned int size ) {
+	bigint_t ( size ) __attribute__ (( may_alias )) *value =
+		( ( void * ) value0 );
+	uint64_t *discard_subtrahend;
+	uint64_t *discard_value;
+	uint64_t discard_subtrahend_i;
+	uint64_t discard_value_i;
+	unsigned int discard_size;
+
+	__asm__ __volatile__ ( "cmp xzr, xzr\n\t" /* set CF */
+			       "\n1:\n\t"
+			       "ldr %3, [%0], #8\n\t"
+			       "ldr %4, [%1]\n\t"
+			       "sbcs %4, %4, %3\n\t"
+			       "str %4, [%1], #8\n\t"
+			       "sub %w2, %w2, #1\n\t"
+			       "cbnz %w2, 1b\n\t"
+			       : "=r" ( discard_subtrahend ),
+				 "=r" ( discard_value ),
+				 "=r" ( discard_size ),
+				 "=r" ( discard_subtrahend_i ),
+				 "=r" ( discard_value_i ),
+				 "+m" ( *value )
+			       : "0" ( subtrahend0 ), "1" ( value0 ),
+				 "2" ( size )
+			       : "cc" );
+}
+
+/**
+ * Rotate big integer left
+ *
+ * @v value0		Element 0 of big integer
+ * @v size		Number of elements
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_rol_raw ( uint64_t *value0, unsigned int size ) {
+	bigint_t ( size ) __attribute__ (( may_alias )) *value =
+		( ( void * ) value0 );
+	uint64_t *discard_value;
+	uint64_t discard_value_i;
+	unsigned int discard_size;
+
+	__asm__ __volatile__ ( "cmn xzr, xzr\n\t" /* clear CF */
+			       "\n1:\n\t"
+			       "ldr %2, [%0]\n\t"
+			       "adcs %2, %2, %2\n\t"
+			       "str %2, [%0], #8\n\t"
+			       "sub %w1, %w1, #1\n\t"
+			       "cbnz %w1, 1b\n\t"
+			       : "=r" ( discard_value ),
+				 "=r" ( discard_size ),
+				 "=r" ( discard_value_i ),
+				 "+m" ( *value )
+			       : "0" ( value0 ), "1" ( size )
+			       : "cc" );
+}
+
+/**
+ * Rotate big integer right
+ *
+ * @v value0		Element 0 of big integer
+ * @v size		Number of elements
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_ror_raw ( uint64_t *value0, unsigned int size ) {
+	bigint_t ( size ) __attribute__ (( may_alias )) *value =
+		( ( void * ) value0 );
+	uint64_t *discard_value;
+	uint64_t discard_value_i;
+	uint64_t discard_value_j;
+	unsigned int discard_size;
+
+	__asm__ __volatile__ ( "mov %3, #0\n\t"
+			       "\n1:\n\t"
+			       "sub %w1, %w1, #1\n\t"
+			       "ldr %2, [%0, %1, lsl #3]\n\t"
+			       "extr %3, %3, %2, #1\n\t"
+			       "str %3, [%0, %1, lsl #3]\n\t"
+			       "mov %3, %2\n\t"
+			       "cbnz %w1, 1b\n\t"
+			       : "=r" ( discard_value ),
+				 "=r" ( discard_size ),
+				 "=r" ( discard_value_i ),
+				 "=r" ( discard_value_j ),
+				 "+m" ( *value )
+			       : "0" ( value0 ), "1" ( size ) );
+}
+
+/**
+ * Test if big integer is equal to zero
+ *
+ * @v value0		Element 0 of big integer
+ * @v size		Number of elements
+ * @ret is_zero		Big integer is equal to zero
+ */
+static inline __attribute__ (( always_inline, pure )) int
+bigint_is_zero_raw ( const uint64_t *value0, unsigned int size ) {
+	const uint64_t *value = value0;
+	uint64_t value_i;
+
+	do {
+		value_i = *(value++);
+		if ( value_i )
+			break;
+	} while ( --size );
+
+	return ( value_i == 0 );
+}
+
+/**
+ * Compare big integers
+ *
+ * @v value0		Element 0 of big integer
+ * @v reference0	Element 0 of reference big integer
+ * @v size		Number of elements
+ * @ret geq		Big integer is greater than or equal to the reference
+ */
+static inline __attribute__ (( always_inline, pure )) int
+bigint_is_geq_raw ( const uint64_t *value0, const uint64_t *reference0,
+		    unsigned int size ) {
+	const uint64_t *value = ( value0 + size );
+	const uint64_t *reference = ( reference0 + size );
+	uint64_t value_i;
+	uint64_t reference_i;
+
+	do {
+		value_i = *(--value);
+		reference_i = *(--reference);
+		if ( value_i != reference_i )
+			break;
+	} while ( --size );
+
+	return ( value_i >= reference_i );
+}
+
+/**
+ * Test if bit is set in big integer
+ *
+ * @v value0		Element 0 of big integer
+ * @v size		Number of elements
+ * @v bit		Bit to test
+ * @ret is_set		Bit is set
+ */
+static inline __attribute__ (( always_inline )) int
+bigint_bit_is_set_raw ( const uint64_t *value0, unsigned int size,
+			unsigned int bit ) {
+	const bigint_t ( size ) __attribute__ (( may_alias )) *value =
+		( ( const void * ) value0 );
+	unsigned int index = ( bit / ( 8 * sizeof ( value->element[0] ) ) );
+	unsigned int subindex = ( bit % ( 8 * sizeof ( value->element[0] ) ) );
+
+	return ( !! ( value->element[index] & ( 1UL << subindex ) ) );
+}
+
+/**
+ * Find highest bit set in big integer
+ *
+ * @v value0		Element 0 of big integer
+ * @v size		Number of elements
+ * @ret max_bit		Highest bit set + 1 (or 0 if no bits set)
+ */
+static inline __attribute__ (( always_inline )) int
+bigint_max_set_bit_raw ( const uint64_t *value0, unsigned int size ) {
+	const uint64_t *value = ( value0 + size );
+	int max_bit = ( 8 * sizeof ( bigint_t ( size ) ) );
+	uint64_t value_i;
+
+	do {
+		value_i = *(--value);
+		max_bit -= ( 64 - fls ( value_i ) );
+		if ( value_i )
+			break;
+	} while ( --size );
+
+	return max_bit;
+}
+
+/**
+ * Grow big integer
+ *
+ * @v source0		Element 0 of source big integer
+ * @v source_size	Number of elements in source big integer
+ * @v dest0		Element 0 of destination big integer
+ * @v dest_size		Number of elements in destination big integer
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_grow_raw ( const uint64_t *source0, unsigned int source_size,
+		  uint64_t *dest0, unsigned int dest_size ) {
+	unsigned int pad_size = ( dest_size - source_size );
+
+	memcpy ( dest0, source0, sizeof ( bigint_t ( source_size ) ) );
+	memset ( ( dest0 + source_size ), 0, sizeof ( bigint_t ( pad_size ) ) );
+}
+
+/**
+ * Shrink big integer
+ *
+ * @v source0		Element 0 of source big integer
+ * @v source_size	Number of elements in source big integer
+ * @v dest0		Element 0 of destination big integer
+ * @v dest_size		Number of elements in destination big integer
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_shrink_raw ( const uint64_t *source0, unsigned int source_size __unused,
+		    uint64_t *dest0, unsigned int dest_size ) {
+
+	memcpy ( dest0, source0, sizeof ( bigint_t ( dest_size ) ) );
+}
+
+/**
+ * Finalise big integer
+ *
+ * @v value0		Element 0 of big integer to finalise
+ * @v size		Number of elements
+ * @v out		Output buffer
+ * @v len		Length of output buffer
+ */
+static inline __attribute__ (( always_inline )) void
+bigint_done_raw ( const uint64_t *value0, unsigned int size __unused,
+		  void *out, size_t len ) {
+	const uint8_t *value_byte = ( ( const void * ) value0 );
+	uint8_t *out_byte = ( out + len );
+
+	/* Copy raw data in reverse order */
+	while ( len-- )
+		*(--out_byte) = *(value_byte++);
+}
+
+extern void bigint_multiply_raw ( const uint64_t *multiplicand0,
+				  const uint64_t *multiplier0,
+				  uint64_t *value0, unsigned int size );
+
+#endif /* _BITS_BIGINT_H */

src/arch/arm64/include/bits/bitops.h

@@ -0,0 +1,100 @@
+#ifndef _BITS_BITOPS_H
+#define _BITS_BITOPS_H
+
+/** @file
+ *
+ * ARM bit operations
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+
+/**
+ * Test and set bit atomically
+ *
+ * @v bit		Bit to set
+ * @v bits		Bit field
+ * @ret old		Old value of bit (zero or non-zero)
+ */
+static inline __attribute__ (( always_inline )) int
+test_and_set_bit ( unsigned int bit, volatile void *bits ) {
+	unsigned int index = ( bit / 64 );
+	unsigned int offset = ( bit % 64 );
+	volatile uint64_t *qword = ( ( ( volatile uint64_t * ) bits ) + index );
+	uint64_t mask = ( 1UL << offset );
+	uint64_t old;
+	uint64_t new;
+	uint32_t flag;
+
+	__asm__ __volatile__ ( "\n1:\n\t"
+			       "ldxr %0, %3\n\t"
+			       "orr %1, %0, %4\n\t"
+			       "stxr %w2, %1, %3\n\t"
+			       "tst %w2, %w2\n\t"
+			       "bne 1b\n\t"
+			       : "=&r" ( old ), "=&r" ( new ), "=&r" ( flag ),
+				 "+Q" ( *qword )
+			       : "r" ( mask )
+			       : "cc" );
+
+	return ( !! ( old & mask ) );
+}
+
+/**
+ * Test and clear bit atomically
+ *
+ * @v bit		Bit to set
+ * @v bits		Bit field
+ * @ret old		Old value of bit (zero or non-zero)
+ */
+static inline __attribute__ (( always_inline )) int
+test_and_clear_bit ( unsigned int bit, volatile void *bits ) {
+	unsigned int index = ( bit / 64 );
+	unsigned int offset = ( bit % 64 );
+	volatile uint64_t *qword = ( ( ( volatile uint64_t * ) bits ) + index );
+	uint64_t mask = ( 1UL << offset );
+	uint64_t old;
+	uint64_t new;
+	uint32_t flag;
+
+	__asm__ __volatile__ ( "\n1:\n\t"
+			       "ldxr %0, %3\n\t"
+			       "bic %1, %0, %4\n\t"
+			       "stxr %w2, %1, %3\n\t"
+			       "tst %w2, %w2\n\t"
+			       "bne 1b\n\t"
+			       : "=&r" ( old ), "=&r" ( new ), "=&r" ( flag ),
+				 "+Q" ( *qword )
+			       : "r" ( mask )
+			       : "cc" );
+
+	return ( !! ( old & mask ) );
+}
+
+/**
+ * Set bit atomically
+ *
+ * @v bit		Bit to set
+ * @v bits		Bit field
+ */
+static inline __attribute__ (( always_inline )) void
+set_bit ( unsigned int bit, volatile void *bits ) {
+
+	test_and_set_bit ( bit, bits );
+}
+
+/**
+ * Clear bit atomically
+ *
+ * @v bit		Bit to set
+ * @v bits		Bit field
+ */
+static inline __attribute__ (( always_inline )) void
+clear_bit ( unsigned int bit, volatile void *bits ) {
+
+	test_and_clear_bit ( bit, bits );
+}
+
+#endif /* _BITS_BITOPS_H */

src/arch/arm64/include/bits/byteswap.h

@@ -0,0 +1,47 @@
+#ifndef _BITS_BYTESWAP_H
+#define _BITS_BYTESWAP_H
+
+/** @file
+ *
+ * Byte-order swapping functions
+ *
+ */
+
+#include <stdint.h>
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+static inline __attribute__ (( always_inline, const )) uint16_t
+__bswap_variable_16 ( uint16_t x ) {
+	__asm__ ( "rev16 %0, %1" : "=r" ( x ) : "r" ( x ) );
+	return x;
+}
+
+static inline __attribute__ (( always_inline )) void
+__bswap_16s ( uint16_t *x ) {
+	*x = __bswap_variable_16 ( *x );
+}
+
+static inline __attribute__ (( always_inline, const )) uint32_t
+__bswap_variable_32 ( uint32_t x ) {
+	__asm__ ( "rev32 %0, %1" : "=r" ( x ) : "r" ( x ) );
+	return x;
+}
+
+static inline __attribute__ (( always_inline )) void
+__bswap_32s ( uint32_t *x ) {
+	*x = __bswap_variable_32 ( *x );
+}
+
+static inline __attribute__ (( always_inline, const )) uint64_t
+__bswap_variable_64 ( uint64_t x ) {
+	__asm__ ( "rev %0, %1" : "=r" ( x ) : "r" ( x ) );
+	return x;
+}
+
+static inline __attribute__ (( always_inline )) void
+__bswap_64s ( uint64_t *x ) {
+	*x = __bswap_variable_64 ( *x );
+}
+
+#endif

src/arch/arm64/include/bits/compiler.h

@@ -0,0 +1,16 @@
+#ifndef _BITS_COMPILER_H
+#define _BITS_COMPILER_H
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/** Dummy relocation type */
+#define RELOC_TYPE_NONE R_AARCH64_NULL
+
+#ifndef ASSEMBLY
+
+#define __asmcall
+#define __libgcc
+
+#endif /* ASSEMBLY */
+
+#endif /*_BITS_COMPILER_H */

src/arch/arm64/include/bits/profile.h

@@ -0,0 +1,28 @@
+#ifndef _BITS_PROFILE_H
+#define _BITS_PROFILE_H
+
+/** @file
+ *
+ * Profiling
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+
+/**
+ * Get profiling timestamp
+ *
+ * @ret timestamp	Timestamp
+ */
+static inline __attribute__ (( always_inline )) uint64_t
+profile_timestamp ( void ) {
+	uint64_t cycles;
+
+	/* Read cycle counter */
+	__asm__ __volatile__ ( "mrs %0, CNTVCT_EL0\n\t" : "=r" ( cycles ) );
+	return cycles;
+}
+
+#endif /* _BITS_PROFILE_H */

src/arch/arm64/include/bits/stdint.h

@@ -0,0 +1,21 @@
+#ifndef _BITS_STDINT_H
+#define _BITS_STDINT_H
+
+typedef __SIZE_TYPE__		size_t;
+typedef signed long		ssize_t;
+typedef signed long		off_t;
+
+typedef unsigned char		uint8_t;
+typedef unsigned short		uint16_t;
+typedef unsigned int		uint32_t;
+typedef unsigned long long	uint64_t;
+
+typedef signed char		int8_t;
+typedef signed short		int16_t;
+typedef signed int		int32_t;
+typedef signed long long	int64_t;
+
+typedef unsigned long		physaddr_t;
+typedef unsigned long		intptr_t;
+
+#endif /* _BITS_STDINT_H */

src/arch/arm64/include/bits/string.h

@@ -0,0 +1,106 @@
+#ifndef BITS_STRING_H
+#define BITS_STRING_H
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/** @file
+ *
+ * String functions
+ *
+ */
+
+extern void arm64_bzero ( void *dest, size_t len );
+extern void arm64_memset ( void *dest, size_t len, int character );
+extern void arm64_memcpy ( void *dest, const void *src, size_t len );
+extern void arm64_memmove_forwards ( void *dest, const void *src, size_t len );
+extern void arm64_memmove_backwards ( void *dest, const void *src, size_t len );
+extern void arm64_memmove ( void *dest, const void *src, size_t len );
+
+/**
+ * Fill memory region
+ *
+ * @v dest		Destination region
+ * @v character		Fill character
+ * @v len		Length
+ * @ret dest		Destination region
+ */
+static inline __attribute__ (( always_inline )) void *
+memset ( void *dest, int character, size_t len ) {
+
+	/* Allow gcc to generate inline "stX xzr" instructions for
+	 * small, constant lengths.
+	 */
+	if ( __builtin_constant_p ( character ) && ( character == 0 ) &&
+	     __builtin_constant_p ( len ) && ( len <= 64 ) ) {
+		__builtin_memset ( dest, 0, len );
+		return dest;
+	}
+
+	/* For zeroing larger or non-constant lengths, use the
+	 * optimised variable-length zeroing code.
+	 */
+	if ( __builtin_constant_p ( character ) && ( character == 0 ) ) {
+		arm64_bzero ( dest, len );
+		return dest;
+	}
+
+	/* Not necessarily zeroing: use basic variable-length code */
+	arm64_memset ( dest, len, character );
+	return dest;
+}
+
+/**
+ * Copy memory region
+ *
+ * @v dest		Destination region
+ * @v src		Source region
+ * @v len		Length
+ * @ret dest		Destination region
+ */
+static inline __attribute__ (( always_inline )) void *
+memcpy ( void *dest, const void *src, size_t len ) {
+
+	/* Allow gcc to generate inline "ldX"/"stX" instructions for
+	 * small, constant lengths.
+	 */
+	if ( __builtin_constant_p ( len ) && ( len <= 64 ) ) {
+		__builtin_memcpy ( dest, src, len );
+		return dest;
+	}
+
+	/* Otherwise, use variable-length code */
+	arm64_memcpy ( dest, src, len );
+	return dest;
+}
+
+/**
+ * Copy (possibly overlapping) memory region
+ *
+ * @v dest		Destination region
+ * @v src		Source region
+ * @v len		Length
+ * @ret dest		Destination region
+ */
+static inline __attribute__ (( always_inline )) void *
+memmove ( void *dest, const void *src, size_t len ) {
+	ssize_t offset = ( dest - src );
+
+	/* If required direction of copy is known at build time, then
+	 * use the appropriate forwards/backwards copy directly.
+	 */
+	if ( __builtin_constant_p ( offset ) ) {
+		if ( offset <= 0 ) {
+			arm64_memmove_forwards ( dest, src, len );
+			return dest;
+		} else {
+			arm64_memmove_backwards ( dest, src, len );
+			return dest;
+		}
+	}
+
+	/* Otherwise, use ambidirectional copy */
+	arm64_memmove ( dest, src, len );
+	return dest;
+}
+
+#endif /* BITS_STRING_H */

src/arch/arm64/include/bits/strings.h

@@ -0,0 +1,69 @@
+#ifndef _BITS_STRINGS_H
+#define _BITS_STRINGS_H
+
+/** @file
+ *
+ * String functions
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/**
+ * Find first (i.e. least significant) set bit
+ *
+ * @v value		Value
+ * @ret lsb		Least significant bit set in value (LSB=1), or zero
+ */
+static inline __attribute__ (( always_inline )) int __ffsll ( long long value ){
+	unsigned long long bits = value;
+	unsigned long long lsb;
+	unsigned int lz;
+
+	/* Extract least significant set bit */
+	lsb = ( bits & -bits );
+
+	/* Count number of leading zeroes before LSB */
+	__asm__ ( "clz %0, %1" : "=r" ( lz ) : "r" ( lsb ) );
+
+	return ( 64 - lz );
+}
+
+/**
+ * Find first (i.e. least significant) set bit
+ *
+ * @v value		Value
+ * @ret lsb		Least significant bit set in value (LSB=1), or zero
+ */
+static inline __attribute__ (( always_inline )) int __ffsl ( long value ) {
+
+	return __ffsll ( value );
+}
+
+/**
+ * Find last (i.e. most significant) set bit
+ *
+ * @v value		Value
+ * @ret msb		Most significant bit set in value (LSB=1), or zero
+ */
+static inline __attribute__ (( always_inline )) int __flsll ( long long value ){
+	unsigned int lz;
+
+	/* Count number of leading zeroes */
+	__asm__ ( "clz %0, %1" : "=r" ( lz ) : "r" ( value ) );
+
+	return ( 64 - lz );
+}
+
+/**
+ * Find last (i.e. most significant) set bit
+ *
+ * @v value		Value
+ * @ret msb		Most significant bit set in value (LSB=1), or zero
+ */
+static inline __attribute__ (( always_inline )) int __flsl ( long value ) {
+
+	return __flsll ( value );
+}
+
+#endif /* _BITS_STRINGS_H */

src/arch/arm64/include/bits/tcpip.h

@@ -0,0 +1,15 @@
+#ifndef _BITS_TCPIP_H
+#define _BITS_TCPIP_H
+
+/** @file
+ *
+ * Transport-network layer interface
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+extern uint16_t tcpip_continue_chksum ( uint16_t sum, const void *data,
+					size_t len );
+
+#endif /* _BITS_TCPIP_H */

src/arch/arm64/include/efi/ipxe/dhcp_arch.h

@@ -0,0 +1,40 @@
+/*
+ * Copyright (C) 2015 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+#ifndef _DHCP_ARCH_H
+#define _DHCP_ARCH_H
+
+/** @file
+ *
+ * Architecture-specific DHCP options
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <ipxe/dhcp.h>
+
+#define DHCP_ARCH_CLIENT_ARCHITECTURE DHCP_CLIENT_ARCHITECTURE_ARM64
+
+#define DHCP_ARCH_CLIENT_NDI 1 /* UNDI */ , 3, 10 /* v3.10 */
+
+#endif

src/arch/arm64/include/gdbmach.h

@@ -0,0 +1,45 @@
+#ifndef GDBMACH_H
+#define GDBMACH_H
+
+/** @file
+ *
+ * GDB architecture specifics
+ *
+ * This file declares functions for manipulating the machine state and
+ * debugging context.
+ *
+ */
+
+#include <stdint.h>
+
+typedef unsigned long gdbreg_t;
+
+/* Register snapshot */
+enum {
+	/* Not yet implemented */
+	GDBMACH_NREGS,
+};
+
+#define GDBMACH_SIZEOF_REGS ( GDBMACH_NREGS * sizeof ( gdbreg_t ) )
+
+static inline void gdbmach_set_pc ( gdbreg_t *regs, gdbreg_t pc ) {
+	/* Not yet implemented */
+	( void ) regs;
+	( void ) pc;
+}
+
+static inline void gdbmach_set_single_step ( gdbreg_t *regs, int step ) {
+	/* Not yet implemented */
+	( void ) regs;
+	( void ) step;
+}
+
+static inline void gdbmach_breakpoint ( void ) {
+	/* Not yet implemented */
+}
+
+extern int gdbmach_set_breakpoint ( int type, unsigned long addr, size_t len,
+				    int enable );
+extern void gdbmach_init ( void );
+
+#endif /* GDBMACH_H */

src/arch/arm64/include/limits.h

@@ -0,0 +1,59 @@
+#ifndef LIMITS_H
+#define LIMITS_H	1
+
+/* Number of bits in a `char' */
+#define CHAR_BIT	8
+
+/* Minimum and maximum values a `signed char' can hold */
+#define SCHAR_MIN	(-128)
+#define SCHAR_MAX	127
+
+/* Maximum value an `unsigned char' can hold. (Minimum is 0.) */
+#define UCHAR_MAX	255
+
+/* Minimum and maximum values a `char' can hold */
+#define CHAR_MIN	SCHAR_MIN
+#define CHAR_MAX	SCHAR_MAX
+
+/* Minimum and maximum values a `signed short int' can hold */
+#define SHRT_MIN	(-32768)
+#define SHRT_MAX	32767
+
+/* Maximum value an `unsigned short' can hold. (Minimum is 0.) */
+#define USHRT_MAX	65535
+
+
+/* Minimum and maximum values a `signed int' can hold */
+#define INT_MIN		(-INT_MAX - 1)
+#define INT_MAX		2147483647
+
+/* Maximum value an `unsigned int' can hold. (Minimum is 0.) */
+#define UINT_MAX	4294967295U
+
+
+/* Minimum and maximum values a `signed int' can hold */
+#define INT_MAX		2147483647
+#define INT_MIN		(-INT_MAX - 1)
+
+
+/* Maximum value an `unsigned int' can hold. (Minimum is 0.) */
+#define UINT_MAX	4294967295U
+
+
+/* Minimum and maximum values a `signed long' can hold */
+#define LONG_MAX	9223372036854775807L
+#define LONG_MIN	(-LONG_MAX - 1L)
+
+/* Maximum value an `unsigned long' can hold. (Minimum is 0.) */
+#define ULONG_MAX	18446744073709551615UL
+
+/* Minimum and maximum values a `signed long long' can hold */
+#define LLONG_MAX	9223372036854775807LL
+#define LLONG_MIN	(-LONG_MAX - 1LL)
+
+
+/* Maximum value an `unsigned long long' can hold. (Minimum is 0.) */
+#define ULLONG_MAX	18446744073709551615ULL
+
+
+#endif /* LIMITS_H */

src/arch/arm64/include/setjmp.h

@@ -0,0 +1,44 @@
+#ifndef _SETJMP_H
+#define _SETJMP_H
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+
+/** A jump buffer */
+typedef struct {
+	/** Saved x19 */
+	uint64_t x19;
+	/** Saved x20 */
+	uint64_t x20;
+	/** Saved x21 */
+	uint64_t x21;
+	/** Saved x22 */
+	uint64_t x22;
+	/** Saved x23 */
+	uint64_t x23;
+	/** Saved x24 */
+	uint64_t x24;
+	/** Saved x25 */
+	uint64_t x25;
+	/** Saved x26 */
+	uint64_t x26;
+	/** Saved x27 */
+	uint64_t x27;
+	/** Saved x28 */
+	uint64_t x28;
+	/** Saved frame pointer (x29) */
+	uint64_t x29;
+	/** Saved link register (x30) */
+	uint64_t x30;
+	/** Saved stack pointer (x31) */
+	uint64_t sp;
+} jmp_buf[1];
+
+extern int __asmcall __attribute__ (( returns_twice ))
+setjmp ( jmp_buf env );
+
+extern void __asmcall __attribute__ (( noreturn ))
+longjmp ( jmp_buf env, int val );
+
+#endif /* _SETJMP_H */

src/arch/i386/include/bits/hyperv.h

@@ -0,0 +1,49 @@
+#ifndef _BITS_HYPERV_H
+#define _BITS_HYPERV_H
+
+/** @file
+ *
+ * Hyper-V interface
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stddef.h>
+#include <stdint.h>
+#include <ipxe/io.h>
+
+/**
+ * Issue hypercall
+ *
+ * @v hv		Hyper-V hypervisor
+ * @v code		Call code
+ * @v in		Input parameters
+ * @v out		Output parameters
+ * @ret status		Status code
+ */
+static inline __attribute__ (( always_inline )) int
+hv_call ( struct hv_hypervisor *hv, unsigned int code, const void *in,
+	  void *out ) {
+	void *hypercall = hv->hypercall;
+	uint32_t in_phys;
+	uint32_t out_phys;
+	uint32_t discard_ecx;
+	uint32_t discard_edx;
+	uint16_t result;
+
+	in_phys = ( ( __builtin_constant_p ( in ) && ( in == NULL ) )
+		       ? 0 : virt_to_phys ( in ) );
+	out_phys = ( ( __builtin_constant_p ( out ) && ( out == NULL ) )
+		       ? 0 : virt_to_phys ( out ) );
+	__asm__ __volatile__ ( "call *%9"
+			       : "=a" ( result ), "=c" ( discard_ecx ),
+				 "=d" ( discard_edx )
+			       : "d" ( 0 ), "a" ( code ),
+				 "b" ( 0 ), "c" ( in_phys ),
+				 "D" ( 0 ), "S" ( out_phys ),
+				 "m" ( hypercall ) );
+	return result;
+}
+
+#endif /* _BITS_HYPERV_H */

src/arch/i386/include/ipxe/msr.h

@@ -0,0 +1,38 @@
+#ifndef _IPXE_MSR_H
+#define _IPXE_MSR_H
+
+/** @file
+ *
+ * Model-specific registers
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/**
+ * Read model-specific register
+ *
+ * @v msr		Model-specific register
+ * @ret value		Value
+ */
+static inline __attribute__ (( always_inline )) uint64_t
+rdmsr ( unsigned int msr ) {
+	uint64_t value;
+
+	__asm__ __volatile__ ( "rdmsr" : "=A" ( value ) : "c" ( msr ) );
+	return value;
+}
+
+/**
+ * Write model-specific register
+ *
+ * @v msr		Model-specific register
+ * @v value		Value
+ */
+static inline __attribute__ (( always_inline )) void
+wrmsr ( unsigned int msr, uint64_t value ) {
+
+	__asm__ __volatile__ ( "wrmsr" : : "c" ( msr ), "A" ( value ) );
+}
+
+#endif /* _IPXE_MSR_H */

src/arch/i386/tests/gdbstub_test.S

@@ -0,0 +1,54 @@
+	.arch i386
+
+	.section ".data", "aw", @progbits
+watch_me:
+	.long 0xfeedbeef
+
+	.section ".text", "ax", @progbits
+	.code32
+gdbstub_test:
+	/* 1. Read registers test */
+	movl    $0xea010203, %eax
+	movl    $0xeb040506, %ebx
+	movl    $0xec070809, %ecx
+	movl    $0xed0a0b0c, %edx
+	movl    $0x510d0e0f, %esi
+	movl    $0xd1102030, %edi
+	int	$3
+
+	/* 2. Write registers test */
+	int	$3
+
+	/* 3. Read memory test */
+	subl	$8, %esp
+	movl	$0x11223344, 4(%esp)
+	movw	$0x5566, 2(%esp)
+	movb	$0x77, (%esp)
+	int	$3
+
+	/* 4. Write memory test */
+	int	$3
+	addl	$8, %esp
+
+	/* 5. Step test */
+	int	$3
+	nop
+
+	/* 6. Access watch test */
+	movl	$0x600d0000, %ecx
+	movl	watch_me, %eax
+	movl	$0xbad00000, %ecx
+	int	$3
+	movl	$0x600d0001, %ecx
+	movl	%eax, watch_me
+	movl	$0xbad00001, %ecx
+	int	$3
+
+	/* 7. Write watch test */
+	movl	$0x600d0002, %ecx
+	movl	%eax, watch_me
+	movl	$0xbad00002, %ecx
+	int	$3
+
+1:
+	jmp	1b

src/arch/i386/tests/gdbstub_test.gdb

@@ -0,0 +1,116 @@
+#!/usr/bin/gdb -x
+# Test suite for GDB remote debugging
+# Run:
+#   make bin/ipxe.hd.tmp
+#   make
+#   gdb
+#   (gdb) target remote :TCPPORT
+#   OR
+#   (gdb) target remote udp:IP:UDPPORT
+#   (gdb) source tests/gdbstub_test.gdb
+
+define ipxe_load_symbols
+	file bin/ipxe.hd.tmp
+end
+
+define ipxe_assert
+	if $arg0 != $arg1
+		echo FAIL $arg2\n
+	else
+		echo PASS $arg2\n
+	end
+end
+
+define ipxe_start_tests
+	jump gdbstub_test
+end
+
+define ipxe_test_regs_read
+	ipxe_assert $eax 0xea010203 "ipxe_test_regs_read eax"
+	ipxe_assert $ebx 0xeb040506 "ipxe_test_regs_read ebx"
+	ipxe_assert $ecx 0xec070809 "ipxe_test_regs_read ecx"
+	ipxe_assert $edx 0xed0a0b0c "ipxe_test_regs_read edx"
+	ipxe_assert $esi 0x510d0e0f "ipxe_test_regs_read esi"
+	ipxe_assert $edi 0xd1102030 "ipxe_test_regs_read edi"
+end
+
+define ipxe_test_regs_write
+	set $eax = 0xea112233
+	set $ebx = 0xeb445566
+	set $ecx = 0xec778899
+	set $edx = 0xedaabbcc
+	set $esi = 0x51ddeeff
+	set $edi = 0xd1010203
+	c
+	ipxe_assert $eax 0xea112233 "ipxe_test_regs_write eax"
+	ipxe_assert $ebx 0xeb445566 "ipxe_test_regs_write ebx"
+	ipxe_assert $ecx 0xec778899 "ipxe_test_regs_write ecx"
+	ipxe_assert $edx 0xedaabbcc "ipxe_test_regs_write edx"
+	ipxe_assert $esi 0x51ddeeff "ipxe_test_regs_write esi"
+	ipxe_assert $edi 0xd1010203 "ipxe_test_regs_write edi"
+
+	# This assumes segment selectors are always 0x10 or 0x8 (for code).
+	ipxe_assert $cs 0x08 "ipxe_test_regs_write cs"
+	ipxe_assert $ds 0x10 "ipxe_test_regs_write ds"
+end
+
+define ipxe_test_mem_read
+	c
+	ipxe_assert ({int}($esp+4)) 0x11223344 "ipxe_test_mem_read int"
+	ipxe_assert ({short}($esp+2)) 0x5566 "ipxe_test_mem_read short"
+	ipxe_assert ({char}($esp)) 0x77 "ipxe_test_mem_read char"
+end
+
+define ipxe_test_mem_write
+	set ({int}($esp+4)) = 0xaabbccdd
+	set ({short}($esp+2)) = 0xeeff
+	set ({char}($esp)) = 0x99
+	c
+	ipxe_assert ({int}($esp+4)) 0xaabbccdd "ipxe_test_mem_write int"
+	ipxe_assert ({short}($esp+2)) (short)0xeeff "ipxe_test_mem_write short"
+	ipxe_assert ({char}($esp)) (char)0x99 "ipxe_test_mem_write char"
+end
+
+define ipxe_test_step
+	c
+	si
+	ipxe_assert ({char}($eip-1)) (char)0x90 "ipxe_test_step" # nop = 0x90
+end
+
+define ipxe_test_awatch
+	awatch watch_me
+
+	c
+	ipxe_assert $ecx 0x600d0000 "ipxe_test_awatch read"
+	if $ecx == 0x600d0000
+		c
+	end
+
+	c
+	ipxe_assert $ecx 0x600d0001 "ipxe_test_awatch write"
+	if $ecx == 0x600d0001
+		c
+	end
+
+	delete
+end
+
+define ipxe_test_watch
+	watch watch_me
+	c
+	ipxe_assert $ecx 0x600d0002 "ipxe_test_watch"
+	if $ecx == 0x600d0002
+		c
+	end
+	delete
+end
+
+ipxe_load_symbols
+ipxe_start_tests
+ipxe_test_regs_read
+ipxe_test_regs_write
+ipxe_test_mem_read
+ipxe_test_mem_write
+ipxe_test_step
+ipxe_test_awatch
+ipxe_test_watch

src/arch/x86/Makefile.pcbios

@@ -0,0 +1,132 @@
+# -*- makefile -*- : Force emacs to use Makefile mode
+
+# BIOS-specific directories containing source files
+#
+SRCDIRS		+= arch/x86/drivers/net
+
+# The i386 linker script
+#
+LDSCRIPT	= arch/x86/scripts/pcbios.lds
+
+# Stop ld from complaining about our customised linker script
+#
+LDFLAGS		+= -N --no-check-sections
+
+# Prefix always starts at address zero
+#
+LDFLAGS		+= --section-start=.prefix=0
+
+# Media types.
+#
+MEDIA		+= rom
+MEDIA		+= mrom
+MEDIA		+= pcirom
+MEDIA		+= isarom
+MEDIA		+= pxe
+MEDIA		+= kpxe
+MEDIA		+= kkpxe
+MEDIA		+= kkkpxe
+MEDIA		+= lkrn
+MEDIA		+= dsk
+MEDIA		+= nbi
+MEDIA		+= hd
+MEDIA		+= raw
+MEDIA		+= exe
+
+# Padding rules
+#
+PAD_rom		= $(PERL) $(PADIMG) --blksize=512 --byte=0xff
+PAD_mrom	= $(PAD_rom)
+PAD_pcirom	= $(PAD_rom)
+PAD_isarom	= $(PAD_rom)
+PAD_dsk		= $(PERL) $(PADIMG) --blksize=512
+PAD_hd		= $(PERL) $(PADIMG) --blksize=32768
+PAD_exe		= $(PERL) $(PADIMG) --blksize=512
+
+# Finalisation rules
+#
+FINALISE_rom	= $(PERL) $(FIXROM)
+FINALISE_mrom	= $(FINALISE_rom)
+FINALISE_pcirom	= $(FINALISE_rom)
+FINALISE_isarom	= $(FINALISE_rom)
+
+# Use $(ROMS) rather than $(DRIVERS) for "allroms", "allmroms", etc.
+#
+LIST_NAME_rom := ROMS
+LIST_NAME_mrom := ROMS
+LIST_NAME_pcirom := ROMS
+LIST_NAME_isarom := ROMS
+
+# Locations of isolinux files
+#
+SYSLINUX_DIR_LIST	:= \
+	/usr/lib/syslinux \
+	/usr/lib/syslinux/bios \
+	/usr/lib/syslinux/modules/bios \
+	/usr/share/syslinux \
+	/usr/share/syslinux/bios \
+	/usr/share/syslinux/modules/bios \
+	/usr/local/share/syslinux \
+	/usr/local/share/syslinux/bios \
+	/usr/local/share/syslinux/modules/bios \
+	/usr/lib/ISOLINUX
+ISOLINUX_BIN_LIST	:= \
+	$(ISOLINUX_BIN) \
+	$(patsubst %,%/isolinux.bin,$(SYSLINUX_DIR_LIST))
+LDLINUX_C32_LIST	:= \
+	$(LDLINUX_C32) \
+	$(patsubst %,%/ldlinux.c32,$(SYSLINUX_DIR_LIST))
+ISOLINUX_BIN	= $(firstword $(wildcard $(ISOLINUX_BIN_LIST)))
+LDLINUX_C32	= $(firstword $(wildcard $(LDLINUX_C32_LIST)))
+
+# rule to make a non-emulation ISO boot image
+NON_AUTO_MEDIA	+= iso
+%iso:	%lkrn util/geniso
+	$(QM)$(ECHO) "  [GENISO] $@"
+	$(Q)ISOLINUX_BIN=$(ISOLINUX_BIN) LDLINUX_C32=$(LDLINUX_C32) \
+	    VERSION="$(VERSION)" bash util/geniso -o $@ $<
+
+# rule to make a floppy emulation ISO boot image
+NON_AUTO_MEDIA	+= liso
+%liso:	%lkrn util/geniso
+	$(QM)$(ECHO) "  [GENISO] $@"
+	$(Q)VERSION="$(VERSION)" bash util/geniso -l -o $@ $<
+
+# rule to make a syslinux floppy image (mountable, bootable)
+NON_AUTO_MEDIA	+= sdsk
+%sdsk:	%lkrn util/gensdsk
+	$(QM)$(ECHO) "  [GENSDSK] $@"
+	$(Q)bash util/gensdsk $@ $<
+
+# rule to write disk images to /dev/fd0
+NON_AUTO_MEDIA	+= fd0
+%fd0 : %dsk
+	$(QM)$(ECHO) "  [DD] $@"
+	$(Q)dd if=$< bs=512 conv=sync of=/dev/fd0
+	$(Q)sync
+
+# Special target for building Master Boot Record binary
+$(BIN)/mbr.tmp : $(BIN)/mbr.o
+	$(QM)$(ECHO) "  [LD] $@"
+	$(Q)$(LD) $(LDFLAGS) -o $@ -e mbr $<
+
+# rule to make a USB disk image
+$(BIN)/usbdisk.tmp : $(BIN)/usbdisk.o
+	$(QM)$(ECHO) "  [LD] $@"
+	$(Q)$(LD) $(LDFLAGS) -o $@ -e mbr $<
+
+NON_AUTO_MEDIA	+= usb
+%usb: $(BIN)/usbdisk.bin %hd
+	$(QM)$(ECHO) "  [FINISH] $@"
+	$(Q)cat $^ > $@
+
+NON_AUTO_MEDIA	+= vhd
+%vhd: %usb
+	$(QM)$(ECHO) "  [FINISH] $@"
+	$(Q)$(QEMUIMG) convert -f raw -O vpc $< $@
+
+# Padded floppy image (e.g. for iLO)
+NON_AUTO_MEDIA += pdsk
+%pdsk : %dsk
+	$(Q)cp $< $@
+	$(Q)$(PADIMG) --blksize=1474560 $@

src/arch/x86/core/basemem_packet.c

@@ -0,0 +1,37 @@
+/*
+ * Copyright (C) 2007 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/**
+ * @file
+ *
+ * Packet buffer in base memory.  Used by various components which
+ * need to pass packets to and from external real-mode code.
+ *
+ */
+
+#include <basemem_packet.h>
+
+#undef basemem_packet
+char __bss16_array ( basemem_packet, [BASEMEM_PACKET_LEN] );

src/arch/x86/core/cachedhcp.c

@@ -0,0 +1,179 @@
+/*
+ * Copyright (C) 2013 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+#include <stdlib.h>
+#include <ipxe/dhcppkt.h>
+#include <ipxe/init.h>
+#include <ipxe/netdevice.h>
+#include <realmode.h>
+#include <pxe_api.h>
+
+/** @file
+ *
+ * Cached DHCP packet
+ *
+ */
+
+/** Cached DHCPACK physical address
+ *
+ * This can be set by the prefix.
+ */
+uint32_t __bss16 ( cached_dhcpack_phys );
+#define cached_dhcpack_phys __use_data16 ( cached_dhcpack_phys )
+
+/** Colour for debug messages */
+#define colour &cached_dhcpack_phys
+
+/** Cached DHCPACK */
+static struct dhcp_packet *cached_dhcpack;
+
+/**
+ * Cached DHCPACK startup function
+ *
+ */
+static void cachedhcp_init ( void ) {
+	struct dhcp_packet *dhcppkt;
+	struct dhcp_packet *tmp;
+	struct dhcphdr *dhcphdr;
+	size_t max_len;
+	size_t len;
+
+	/* Do nothing if no cached DHCPACK is present */
+	if ( ! cached_dhcpack_phys ) {
+		DBGC ( colour, "CACHEDHCP found no cached DHCPACK\n" );
+		return;
+	}
+
+	/* No reliable way to determine length before parsing packet;
+	 * start by assuming maximum length permitted by PXE.
+	 */
+	max_len = sizeof ( BOOTPLAYER_t );
+
+	/* Allocate and populate DHCP packet */
+	dhcppkt = zalloc ( sizeof ( *dhcppkt ) + max_len );
+	if ( ! dhcppkt ) {
+		DBGC ( colour, "CACHEDHCP could not allocate copy\n" );
+		return;
+	}
+	dhcphdr = ( ( ( void * ) dhcppkt ) + sizeof ( *dhcppkt ) );
+	copy_from_user ( dhcphdr, phys_to_user ( cached_dhcpack_phys ), 0,
+			 max_len );
+	dhcppkt_init ( dhcppkt, dhcphdr, max_len );
+
+	/* Shrink packet to required length.  If reallocation fails,
+	 * just continue to use the original packet and waste the
+	 * unused space.
+	 */
+	len = dhcppkt_len ( dhcppkt );
+	assert ( len <= max_len );
+	tmp = realloc ( dhcppkt, ( sizeof ( *dhcppkt ) + len ) );
+	if ( tmp )
+		dhcppkt = tmp;
+
+	/* Reinitialise packet at new address */
+	dhcphdr = ( ( ( void * ) dhcppkt ) + sizeof ( *dhcppkt ) );
+	dhcppkt_init ( dhcppkt, dhcphdr, len );
+
+	/* Store as cached DHCPACK, and mark original copy as consumed */
+	DBGC ( colour, "CACHEDHCP found cached DHCPACK at %08x+%zx\n",
+	       cached_dhcpack_phys, len );
+	cached_dhcpack = dhcppkt;
+	cached_dhcpack_phys = 0;
+}
+
+/**
+ * Cached DHCPACK startup function
+ *
+ */
+static void cachedhcp_startup ( void ) {
+
+	/* If cached DHCP packet was not claimed by any network device
+	 * during startup, then free it.
+	 */
+	if ( cached_dhcpack ) {
+		DBGC ( colour, "CACHEDHCP freeing unclaimed cached DHCPACK\n" );
+		dhcppkt_put ( cached_dhcpack );
+		cached_dhcpack = NULL;
+	}
+}
+
+/** Cached DHCPACK initialisation function */
+struct init_fn cachedhcp_init_fn __init_fn ( INIT_NORMAL ) = {
+	.initialise = cachedhcp_init,
+};
+
+/** Cached DHCPACK startup function */
+struct startup_fn cachedhcp_startup_fn __startup_fn ( STARTUP_LATE ) = {
+	.name = "cachedhcp",
+	.startup = cachedhcp_startup,
+};
+
+/**
+ * Apply cached DHCPACK to network device, if applicable
+ *
+ * @v netdev		Network device
+ * @ret rc		Return status code
+ */
+static int cachedhcp_probe ( struct net_device *netdev ) {
+	struct ll_protocol *ll_protocol = netdev->ll_protocol;
+	int rc;
+
+	/* Do nothing unless we have a cached DHCPACK */
+	if ( ! cached_dhcpack )
+		return 0;
+
+	/* Do nothing unless cached DHCPACK's MAC address matches this
+	 * network device.
+	 */
+	if ( memcmp ( netdev->ll_addr, cached_dhcpack->dhcphdr->chaddr,
+		      ll_protocol->ll_addr_len ) != 0 ) {
+		DBGC ( colour, "CACHEDHCP cached DHCPACK does not match %s\n",
+		       netdev->name );
+		return 0;
+	}
+	DBGC ( colour, "CACHEDHCP cached DHCPACK is for %s\n", netdev->name );
+
+	/* Register as DHCP settings for this network device */
+	if ( ( rc = register_settings ( &cached_dhcpack->settings,
+					netdev_settings ( netdev ),
+					DHCP_SETTINGS_NAME ) ) != 0 ) {
+		DBGC ( colour, "CACHEDHCP could not register settings: %s\n",
+		       strerror ( rc ) );
+		return rc;
+	}
+
+	/* Claim cached DHCPACK */
+	dhcppkt_put ( cached_dhcpack );
+	cached_dhcpack = NULL;
+
+	return 0;
+}
+
+/** Cached DHCP packet network device driver */
+struct net_driver cachedhcp_driver __net_driver = {
+	.name = "cachedhcp",
+	.probe = cachedhcp_probe,
+};

src/arch/x86/core/dumpregs.c

@@ -0,0 +1,20 @@
+#include <stdio.h>
+#include <realmode.h>
+
+void __asmcall _dump_regs ( struct i386_all_regs *ix86 ) {
+
+	__asm__ __volatile__ (
+		TEXT16_CODE ( ".globl dump_regs\n\t"
+			      "\ndump_regs:\n\t"
+			      VIRT_CALL ( _dump_regs )
+			      "ret\n\t" ) : );
+
+	printf ( "EAX=%08x EBX=%08x ECX=%08x EDX=%08x\n"
+		 "ESI=%08x EDI=%08x EBP=%08x ESP=%08x\n"
+		 "CS=%04x SS=%04x DS=%04x ES=%04x FS=%04x GS=%04x\n",
+		 ix86->regs.eax, ix86->regs.ebx, ix86->regs.ecx,
+		 ix86->regs.edx, ix86->regs.esi, ix86->regs.edi,
+		 ix86->regs.ebp, ix86->regs.esp,
+		 ix86->segs.cs, ix86->segs.ss, ix86->segs.ds,
+		 ix86->segs.es, ix86->segs.fs, ix86->segs.gs );
+}

src/arch/x86/core/gdbmach.c

@@ -0,0 +1,251 @@
+/*
+ * Copyright (C) 2008 Stefan Hajnoczi <stefanha@gmail.com>.
+ * Copyright (C) 2016 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stddef.h>
+#include <stdio.h>
+#include <errno.h>
+#include <assert.h>
+#include <ipxe/uaccess.h>
+#include <ipxe/gdbstub.h>
+#include <librm.h>
+#include <gdbmach.h>
+
+/** @file
+ *
+ * GDB architecture-specific bits for x86
+ *
+ */
+
+/** Number of hardware breakpoints */
+#define NUM_HWBP 4
+
+/** Debug register 7: Global breakpoint enable */
+#define DR7_G( bp ) ( 2 << ( 2 * (bp) ) )
+
+/** Debug register 7: Global exact breakpoint enable */
+#define DR7_GE ( 1 << 9 )
+
+/** Debug register 7: Break on data writes */
+#define DR7_RWLEN_WRITE 0x11110000
+
+/** Debug register 7: Break on data access */
+#define DR7_RWLEN_ACCESS 0x33330000
+
+/** Debug register 7: One-byte length */
+#define DR7_RWLEN_1 0x00000000
+
+/** Debug register 7: Two-byte length */
+#define DR7_RWLEN_2 0x44440000
+
+/** Debug register 7: Four-byte length */
+#define DR7_RWLEN_4 0xcccc0000
+
+/** Debug register 7: Eight-byte length */
+#define DR7_RWLEN_8 0x88880000
+
+/** Debug register 7: Breakpoint R/W and length mask */
+#define DR7_RWLEN_MASK( bp ) ( 0xf0000 << ( 4 * (bp) ) )
+
+/** Hardware breakpoint addresses (debug registers 0-3) */
+static unsigned long dr[NUM_HWBP];
+
+/** Active value of debug register 7 */
+static unsigned long dr7 = DR7_GE;
+
+/**
+ * Update debug registers
+ *
+ */
+static void gdbmach_update ( void ) {
+
+	/* Set debug registers */
+	__asm__ __volatile__ ( "mov %0, %%dr0" : : "r" ( dr[0] ) );
+	__asm__ __volatile__ ( "mov %0, %%dr1" : : "r" ( dr[1] ) );
+	__asm__ __volatile__ ( "mov %0, %%dr2" : : "r" ( dr[2] ) );
+	__asm__ __volatile__ ( "mov %0, %%dr3" : : "r" ( dr[3] ) );
+	__asm__ __volatile__ ( "mov %0, %%dr7" : : "r" ( dr7 ) );
+}
+
+/**
+ * Find reusable or available hardware breakpoint
+ *
+ * @v addr		Linear address
+ * @v rwlen		Control bits
+ * @ret bp		Hardware breakpoint, or negative error
+ */
+static int gdbmach_find ( unsigned long addr, unsigned int rwlen ) {
+	unsigned int i;
+	int bp = -ENOENT;
+
+	/* Look for a reusable or available breakpoint */
+	for ( i = 0 ; i < NUM_HWBP ; i++ ) {
+
+		/* If breakpoint is not enabled, then it is available */
+		if ( ! ( dr7 & DR7_G ( i ) ) ) {
+			bp = i;
+			continue;
+		}
+
+		/* If breakpoint is enabled and has the same address
+		 * and control bits, then reuse it.
+		 */
+		if ( ( dr[i] == addr ) &&
+		     ( ( ( dr7 ^ rwlen ) & DR7_RWLEN_MASK ( i ) ) == 0 ) ) {
+			bp = i;
+			break;
+		}
+	}
+
+	return bp;
+}
+
+/**
+ * Set hardware breakpoint
+ *
+ * @v type		GDB breakpoint type
+ * @v addr		Virtual address
+ * @v len		Length
+ * @v enable		Enable (not disable) breakpoint
+ * @ret rc		Return status code
+ */
+int gdbmach_set_breakpoint ( int type, unsigned long addr, size_t len,
+			     int enable ) {
+	unsigned int rwlen;
+	unsigned long mask;
+	int bp;
+
+	/* Parse breakpoint type */
+	switch ( type ) {
+	case GDBMACH_WATCH:
+		rwlen = DR7_RWLEN_WRITE;
+		break;
+	case GDBMACH_AWATCH:
+		rwlen = DR7_RWLEN_ACCESS;
+		break;
+	default:
+		return -ENOTSUP;
+	}
+
+	/* Parse breakpoint length */
+	switch ( len ) {
+	case 1:
+		rwlen |= DR7_RWLEN_1;
+		break;
+	case 2:
+		rwlen |= DR7_RWLEN_2;
+		break;
+	case 4:
+		rwlen |= DR7_RWLEN_4;
+		break;
+	case 8:
+		rwlen |= DR7_RWLEN_8;
+		break;
+	default:
+		return -ENOTSUP;
+	}
+
+	/* Convert to linear address */
+	if ( sizeof ( physaddr_t ) <= sizeof ( uint32_t ) )
+		addr = virt_to_phys ( ( void * ) addr );
+
+	/* Find reusable or available hardware breakpoint */
+	bp = gdbmach_find ( addr, rwlen );
+	if ( bp < 0 )
+		return ( enable ? -ENOBUFS : 0 );
+
+	/* Configure this breakpoint */
+	DBGC ( &dr[0], "GDB bp %d at %p+%zx type %d (%sabled)\n",
+	       bp, ( ( void * ) addr ), len, type, ( enable ? "en" : "dis" ) );
+	dr[bp] = addr;
+	mask = DR7_RWLEN_MASK ( bp );
+	dr7 = ( ( dr7 & ~mask ) | ( rwlen & mask ) );
+	mask = DR7_G ( bp );
+	dr7 &= ~mask;
+	if ( enable )
+		dr7 |= mask;
+
+	/* Update debug registers */
+	gdbmach_update();
+
+	return 0;
+}
+
+/**
+ * Handle exception
+ *
+ * @v signo		GDB signal number
+ * @v regs		Register dump
+ */
+__asmcall void gdbmach_handler ( int signo, gdbreg_t *regs ) {
+	unsigned long dr7_disabled = DR7_GE;
+	unsigned long dr6_clear = 0;
+
+	/* Temporarily disable breakpoints */
+	__asm__ __volatile__ ( "mov %0, %%dr7\n" : : "r" ( dr7_disabled ) );
+
+	/* Handle exception */
+	DBGC ( &dr[0], "GDB signal %d\n", signo );
+	DBGC2_HDA ( &dr[0], 0, regs, ( GDBMACH_NREGS * sizeof ( *regs ) ) );
+	gdbstub_handler ( signo, regs );
+	DBGC ( &dr[0], "GDB signal %d returning\n", signo );
+	DBGC2_HDA ( &dr[0], 0, regs, ( GDBMACH_NREGS * sizeof ( *regs ) ) );
+
+	/* Clear breakpoint status register */
+	__asm__ __volatile__ ( "mov %0, %%dr6\n" : : "r" ( dr6_clear ) );
+
+	/* Re-enable breakpoints */
+	__asm__ __volatile__ ( "mov %0, %%dr7\n" : : "r" ( dr7 ) );
+}
+
+/**
+ * CPU exception vectors
+ *
+ * Note that we cannot intercept anything from INT8 (double fault)
+ * upwards, since these overlap by default with IRQ0-7.
+ */
+static void * gdbmach_vectors[] = {
+	gdbmach_sigfpe,		/* Divide by zero */
+	gdbmach_sigtrap,	/* Debug trap */
+	NULL,			/* Non-maskable interrupt */
+	gdbmach_sigtrap,	/* Breakpoint */
+	gdbmach_sigstkflt,	/* Overflow */
+	gdbmach_sigstkflt,	/* Bound range exceeded */
+	gdbmach_sigill,		/* Invalid opcode */
+};
+
+/**
+ * Initialise GDB
+ */
+void gdbmach_init ( void ) {
+	unsigned int i;
+
+	/* Hook CPU exception vectors */
+	for ( i = 0 ; i < ( sizeof ( gdbmach_vectors ) /
+			    sizeof ( gdbmach_vectors[0] ) ) ; i++ ) {
+		if ( gdbmach_vectors[i] )
+			set_interrupt_vector ( i, gdbmach_vectors[i] );
+	}
+}

src/arch/x86/core/patch_cf.S

@@ -0,0 +1,42 @@
+/*
+ * Copyright (C) 2009 H. Peter Anvin <hpa@zytor.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL )
+
+	.text
+	.arch i386
+	.code16
+
+/****************************************************************************
+ * Set/clear CF on the stack as appropriate, assumes stack is as it should
+ * be immediately before IRET
+ ****************************************************************************
+ */
+	.section ".text16", "ax", @progbits
+	.globl patch_cf
+patch_cf:
+	pushw	%bp
+	movw	%sp, %bp
+	setc	8(%bp)	/* Set/reset CF; clears PF, AF, ZF, SF */
+	popw	%bp
+	ret
+	.size patch_cf, . - patch_cf

src/arch/x86/core/pci_autoboot.c

@@ -0,0 +1,48 @@
+/*
+ * Copyright (C) 2014 Red Hat Inc.
+ *	Alex Williamson <alex.williamson@redhat.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+#include <ipxe/device.h>
+#include <ipxe/init.h>
+#include <realmode.h>
+#include <usr/autoboot.h>
+
+uint16_t __bss16 ( autoboot_busdevfn );
+#define autoboot_busdevfn __use_data16 ( autoboot_busdevfn )
+
+/**
+ * Initialise PCI autoboot device
+ */
+static void pci_autoboot_init ( void ) {
+
+	if ( autoboot_busdevfn )
+		set_autoboot_busloc ( BUS_TYPE_PCI, autoboot_busdevfn );
+}
+
+/** PCI autoboot device initialisation function */
+struct init_fn pci_autoboot_init_fn __init_fn ( INIT_NORMAL ) = {
+	.initialise = pci_autoboot_init,
+};

src/arch/x86/core/pic8259.c

@@ -0,0 +1,67 @@
+/*
+ * Copyright (C) 2007 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER );
+
+#include <ipxe/io.h>
+#include <pic8259.h>
+
+/** @file
+ *
+ * Minimal support for the 8259 Programmable Interrupt Controller
+ *
+ */
+
+/**
+ * Send non-specific EOI(s)
+ *
+ * @v irq		IRQ number
+ *
+ * This seems to be inherently unsafe.
+ */
+static inline void send_nonspecific_eoi ( unsigned int irq ) {
+	DBG ( "Sending non-specific EOI for IRQ %d\n", irq );
+	if ( irq >= IRQ_PIC_CUTOFF ) {
+		outb ( ICR_EOI_NON_SPECIFIC, PIC2_ICR );
+	}		
+	outb ( ICR_EOI_NON_SPECIFIC, PIC1_ICR );
+}
+
+/**
+ * Send specific EOI(s)
+ *
+ * @v irq		IRQ number
+ */
+static inline void send_specific_eoi ( unsigned int irq ) {
+	DBG ( "Sending specific EOI for IRQ %d\n", irq );
+	if ( irq >= IRQ_PIC_CUTOFF ) {
+		outb ( ( ICR_EOI_SPECIFIC | ICR_VALUE ( CHAINED_IRQ ) ),
+		       ICR_REG ( CHAINED_IRQ ) );
+	}
+	outb ( ( ICR_EOI_SPECIFIC | ICR_VALUE ( irq ) ), ICR_REG ( irq ) );
+}
+
+/**
+ * Send End-Of-Interrupt to the PIC
+ *
+ * @v irq		IRQ number
+ */
+void send_eoi ( unsigned int irq ) {
+	send_specific_eoi ( irq );
+}

src/arch/x86/core/pit8254.c

@@ -0,0 +1,70 @@
+/*
+ * Copyright (C) 2015 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <assert.h>
+#include <ipxe/io.h>
+#include <ipxe/pit8254.h>
+
+/** @file
+ *
+ * 8254 Programmable Interval Timer
+ *
+ */
+
+/**
+ * Delay for a fixed number of timer ticks using the speaker channel
+ *
+ * @v ticks		Number of timer ticks for which to delay
+ */
+void pit8254_speaker_delay ( unsigned int ticks ) {
+	uint8_t spkr;
+	uint8_t cmd;
+	uint8_t low;
+	uint8_t high;
+
+	/* Sanity check */
+	assert ( ticks <= 0xffff );
+
+	/* Disable speaker, set speaker channel gate input high */
+	spkr = inb ( PIT8254_SPKR );
+	spkr &= ~PIT8254_SPKR_ENABLE;
+	spkr |= PIT8254_SPKR_GATE;
+	outb ( spkr, PIT8254_SPKR );
+
+	/* Program speaker channel to "interrupt" on terminal count */
+	cmd = ( PIT8254_CMD_CHANNEL ( PIT8254_CH_SPKR ) |
+		PIT8254_CMD_ACCESS_LOHI | PIT8254_CMD_OP_TERMINAL |
+		PIT8254_CMD_BINARY );
+	low = ( ( ticks >> 0 ) & 0xff );
+	high = ( ( ticks >> 8 ) & 0xff );
+	outb ( cmd, PIT8254_CMD );
+	outb ( low, PIT8254_DATA ( PIT8254_CH_SPKR ) );
+	outb ( high, PIT8254_DATA ( PIT8254_CH_SPKR ) );
+
+	/* Wait for channel to "interrupt" */
+	do {
+		spkr = inb ( PIT8254_SPKR );
+	} while ( ! ( spkr & PIT8254_SPKR_OUT ) );
+}

src/arch/x86/core/rdtsc_timer.c

@@ -0,0 +1,177 @@
+/*
+ * Copyright (C) 2008 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/** @file
+ *
+ * RDTSC timer
+ *
+ */
+
+#include <string.h>
+#include <errno.h>
+#include <ipxe/timer.h>
+#include <ipxe/cpuid.h>
+#include <ipxe/pit8254.h>
+
+/** Number of microseconds to use for TSC calibration */
+#define TSC_CALIBRATE_US 1024
+
+/** TSC increment per microsecond */
+static unsigned long tsc_per_us;
+
+/** Minimum resolution for scaled TSC timer */
+#define TSC_SCALED_HZ 32
+
+/** TSC scale (expressed as a bit shift)
+ *
+ * We use this to avoid the need for 64-bit divsion on 32-bit systems.
+ */
+static unsigned int tsc_scale;
+
+/** Number of timer ticks per scaled TSC increment */
+static unsigned long ticks_per_scaled_tsc;
+
+/** Colour for debug messages */
+#define colour &tsc_per_us
+
+/**
+ * Get raw TSC value
+ *
+ * @ret tsc		Raw TSC value
+ */
+static inline __always_inline unsigned long rdtsc_raw ( void ) {
+	unsigned long raw;
+
+	__asm__ __volatile__ ( "rdtsc\n\t" : "=a" ( raw ) : : "edx" );
+	return raw;
+}
+
+/**
+ * Get TSC value, shifted to avoid rollover within a realistic timescale
+ *
+ * @ret tsc		Scaled TSC value
+ */
+static inline __always_inline unsigned long rdtsc_scaled ( void ) {
+	unsigned long scaled;
+
+	__asm__ __volatile__ ( "rdtsc\n\t"
+			       "shrdl %b1, %%edx, %%eax\n\t"
+			       : "=a" ( scaled ) : "c" ( tsc_scale ) : "edx" );
+	return scaled;
+}
+
+/**
+ * Get current system time in ticks
+ *
+ * @ret ticks		Current time, in ticks
+ */
+static unsigned long rdtsc_currticks ( void ) {
+	unsigned long scaled;
+
+	scaled = rdtsc_scaled();
+	return ( scaled * ticks_per_scaled_tsc );
+}
+
+/**
+ * Delay for a fixed number of microseconds
+ *
+ * @v usecs		Number of microseconds for which to delay
+ */
+static void rdtsc_udelay ( unsigned long usecs ) {
+	unsigned long start;
+	unsigned long elapsed;
+	unsigned long threshold;
+
+	start = rdtsc_raw();
+	threshold = ( usecs * tsc_per_us );
+	do {
+		elapsed = ( rdtsc_raw() - start );
+	} while ( elapsed < threshold );
+}
+
+/**
+ * Probe RDTSC timer
+ *
+ * @ret rc		Return status code
+ */
+static int rdtsc_probe ( void ) {
+	unsigned long before;
+	unsigned long after;
+	unsigned long elapsed;
+	uint32_t apm;
+	uint32_t discard_a;
+	uint32_t discard_b;
+	uint32_t discard_c;
+	int rc;
+
+	/* Check that TSC is invariant */
+	if ( ( rc = cpuid_supported ( CPUID_APM ) ) != 0 ) {
+		DBGC ( colour, "RDTSC cannot determine APM features: %s\n",
+		       strerror ( rc ) );
+		return rc;
+	}
+	cpuid ( CPUID_APM, 0, &discard_a, &discard_b, &discard_c, &apm );
+	if ( ! ( apm & CPUID_APM_EDX_TSC_INVARIANT ) ) {
+		DBGC ( colour, "RDTSC has non-invariant TSC (%#08x)\n",
+		       apm );
+		return -ENOTTY;
+	}
+
+	/* Calibrate udelay() timer via 8254 PIT */
+	before = rdtsc_raw();
+	pit8254_udelay ( TSC_CALIBRATE_US );
+	after = rdtsc_raw();
+	elapsed = ( after - before );
+	tsc_per_us = ( elapsed / TSC_CALIBRATE_US );
+	if ( ! tsc_per_us ) {
+		DBGC ( colour, "RDTSC has zero TSC per microsecond\n" );
+		return -EIO;
+	}
+
+	/* Calibrate currticks() scaling factor */
+	tsc_scale = 31;
+	ticks_per_scaled_tsc = ( ( 1UL << tsc_scale ) /
+				 ( tsc_per_us * ( 1000000 / TICKS_PER_SEC ) ) );
+	while ( ticks_per_scaled_tsc > ( TICKS_PER_SEC / TSC_SCALED_HZ ) ) {
+		tsc_scale--;
+		ticks_per_scaled_tsc >>= 1;
+	}
+	DBGC ( colour, "RDTSC has %ld tsc per us, %ld ticks per 2^%d tsc\n",
+	       tsc_per_us, ticks_per_scaled_tsc, tsc_scale );
+	if ( ! ticks_per_scaled_tsc ) {
+		DBGC ( colour, "RDTSC has zero ticks per TSC\n" );
+		return -EIO;
+	}
+
+	return 0;
+}
+
+/** RDTSC timer */
+struct timer rdtsc_timer __timer ( TIMER_PREFERRED ) = {
+	.name = "rdtsc",
+	.probe = rdtsc_probe,
+	.currticks = rdtsc_currticks,
+	.udelay = rdtsc_udelay,
+};

src/arch/x86/core/relocate.c

@@ -0,0 +1,136 @@
+#include <ipxe/io.h>
+#include <registers.h>
+
+/*
+ * Originally by Eric Biederman
+ *
+ * Heavily modified by Michael Brown 
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/* Linker symbols */
+extern char _textdata[];
+extern char _etextdata[];
+
+/* within 1MB of 4GB is too close. 
+ * MAX_ADDR is the maximum address we can easily do DMA to.
+ *
+ * Not sure where this constraint comes from, but kept it from Eric's
+ * old code - mcb30
+ */
+#define MAX_ADDR (0xfff00000UL)
+
+/* Preserve alignment to a 4kB page
+ *
+ * Required for x86_64, and doesn't hurt for i386.
+ */
+#define ALIGN 4096
+
+/**
+ * Relocate iPXE
+ *
+ * @v ebp		Maximum address to use for relocation
+ * @ret esi		Current physical address
+ * @ret edi		New physical address
+ * @ret ecx		Length to copy
+ *
+ * This finds a suitable location for iPXE near the top of 32-bit
+ * address space, and returns the physical address of the new location
+ * to the prefix in %edi.
+ */
+__asmcall void relocate ( struct i386_all_regs *ix86 ) {
+	struct memory_map memmap;
+	uint32_t start, end, size, padded_size, max;
+	uint32_t new_start, new_end;
+	unsigned i;
+
+	/* Get memory map and current location */
+	get_memmap ( &memmap );
+	start = virt_to_phys ( _textdata );
+	end = virt_to_phys ( _etextdata );
+	size = ( end - start );
+	padded_size = ( size + ALIGN - 1 );
+
+	DBG ( "Relocate: currently at [%x,%x)\n"
+	      "...need %x bytes for %d-byte alignment\n",
+	      start, end, padded_size, ALIGN );
+
+	/* Determine maximum usable address */
+	max = MAX_ADDR;
+	if ( ix86->regs.ebp < max ) {
+		max = ix86->regs.ebp;
+		DBG ( "Limiting relocation to [0,%x)\n", max );
+	}
+
+	/* Walk through the memory map and find the highest address
+	 * below 4GB that iPXE will fit into.
+	 */
+	new_end = end;
+	for ( i = 0 ; i < memmap.count ; i++ ) {
+		struct memory_region *region = &memmap.regions[i];
+		uint32_t r_start, r_end;
+
+		DBG ( "Considering [%llx,%llx)\n", region->start, region->end);
+		
+		/* Truncate block to maximum address.  This will be
+		 * less than 4GB, which means that we can get away
+		 * with using just 32-bit arithmetic after this stage.
+		 */
+		if ( region->start > max ) {
+			DBG ( "...starts after max=%x\n", max );
+			continue;
+		}
+		r_start = region->start;
+		if ( region->end > max ) {
+			DBG ( "...end truncated to max=%x\n", max );
+			r_end = max;
+		} else {
+			r_end = region->end;
+		}
+		DBG ( "...usable portion is [%x,%x)\n", r_start, r_end );
+
+		/* If we have rounded down r_end below r_ start, skip
+		 * this block.
+		 */
+		if ( r_end < r_start ) {
+			DBG ( "...truncated to negative size\n" );
+			continue;
+		}
+
+		/* Check that there is enough space to fit in iPXE */
+		if ( ( r_end - r_start ) < size ) {
+			DBG ( "...too small (need %x bytes)\n", size );
+			continue;
+		}
+
+		/* If the start address of the iPXE we would
+		 * place in this block is higher than the end address
+		 * of the current highest block, use this block.
+		 *
+		 * Note that this avoids overlaps with the current
+		 * iPXE, as well as choosing the highest of all viable
+		 * blocks.
+		 */
+		if ( ( r_end - size ) > new_end ) {
+			new_end = r_end;
+			DBG ( "...new best block found.\n" );
+		}
+	}
+
+	/* Calculate new location of iPXE, and align it to the
+	 * required alignemnt.
+	 */
+	new_start = new_end - padded_size;
+	new_start += ( ( start - new_start ) & ( ALIGN - 1 ) );
+	new_end = new_start + size;
+
+	DBG ( "Relocating from [%x,%x) to [%x,%x)\n",
+	      start, end, new_start, new_end );
+	
+	/* Let prefix know what to copy */
+	ix86->regs.esi = start;
+	ix86->regs.edi = new_start;
+	ix86->regs.ecx = size;
+}

src/arch/x86/core/runtime.c

@@ -0,0 +1,270 @@
+/*
+ * Copyright (C) 2011 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/** @file
+ *
+ * Command line and initrd passed to iPXE at runtime
+ *
+ */
+
+#include <stddef.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <ctype.h>
+#include <errno.h>
+#include <assert.h>
+#include <ipxe/init.h>
+#include <ipxe/image.h>
+#include <ipxe/script.h>
+#include <ipxe/umalloc.h>
+#include <realmode.h>
+
+/** Command line physical address
+ *
+ * This can be set by the prefix.
+ */
+uint32_t __bss16 ( cmdline_phys );
+#define cmdline_phys __use_data16 ( cmdline_phys )
+
+/** initrd physical address
+ *
+ * This can be set by the prefix.
+ */
+uint32_t __bss16 ( initrd_phys );
+#define initrd_phys __use_data16 ( initrd_phys )
+
+/** initrd length
+ *
+ * This can be set by the prefix.
+ */
+uint32_t __bss16 ( initrd_len );
+#define initrd_len __use_data16 ( initrd_len )
+
+/** Internal copy of the command line */
+static char *cmdline_copy;
+
+/** Free command line image */
+static void cmdline_image_free ( struct refcnt *refcnt ) {
+	struct image *image = container_of ( refcnt, struct image, refcnt );
+
+	DBGC ( image, "RUNTIME freeing command line\n" );
+	free ( cmdline_copy );
+}
+
+/** Embedded script representing the command line */
+static struct image cmdline_image = {
+	.refcnt = REF_INIT ( cmdline_image_free ),
+	.name = "<CMDLINE>",
+	.type = &script_image_type,
+};
+
+/** Colour for debug messages */
+#define colour &cmdline_image
+
+/**
+ * Strip unwanted cruft from command line
+ *
+ * @v cmdline		Command line
+ * @v cruft		Initial substring of cruft to strip
+ */
+static void cmdline_strip ( char *cmdline, const char *cruft ) {
+	char *strip;
+	char *strip_end;
+
+	/* Find unwanted cruft, if present */
+	if ( ! ( strip = strstr ( cmdline, cruft ) ) )
+		return;
+
+	/* Strip unwanted cruft */
+	strip_end = strchr ( strip, ' ' );
+	if ( strip_end ) {
+		*strip_end = '\0';
+		DBGC ( colour, "RUNTIME stripping \"%s\"\n", strip );
+		strcpy ( strip, ( strip_end + 1 ) );
+	} else {
+		DBGC ( colour, "RUNTIME stripping \"%s\"\n", strip );
+		*strip = '\0';
+	}
+}
+
+/**
+ * Initialise command line
+ *
+ * @ret rc		Return status code
+ */
+static int cmdline_init ( void ) {
+	userptr_t cmdline_user;
+	char *cmdline;
+	size_t len;
+	int rc;
+
+	/* Do nothing if no command line was specified */
+	if ( ! cmdline_phys ) {
+		DBGC ( colour, "RUNTIME found no command line\n" );
+		return 0;
+	}
+	cmdline_user = phys_to_user ( cmdline_phys );
+	len = ( strlen_user ( cmdline_user, 0 ) + 1 /* NUL */ );
+
+	/* Allocate and copy command line */
+	cmdline_copy = malloc ( len );
+	if ( ! cmdline_copy ) {
+		DBGC ( colour, "RUNTIME could not allocate %zd bytes for "
+		       "command line\n", len );
+		rc = -ENOMEM;
+		goto err_alloc_cmdline_copy;
+	}
+	cmdline = cmdline_copy;
+	copy_from_user ( cmdline, cmdline_user, 0, len );
+	DBGC ( colour, "RUNTIME found command line \"%s\" at %08x\n",
+	       cmdline, cmdline_phys );
+
+	/* Mark command line as consumed */
+	cmdline_phys = 0;
+
+	/* Strip unwanted cruft from the command line */
+	cmdline_strip ( cmdline, "BOOT_IMAGE=" );
+	cmdline_strip ( cmdline, "initrd=" );
+	while ( isspace ( *cmdline ) )
+		cmdline++;
+	DBGC ( colour, "RUNTIME using command line \"%s\"\n", cmdline );
+
+	/* Prepare and register image */
+	cmdline_image.data = virt_to_user ( cmdline );
+	cmdline_image.len = strlen ( cmdline );
+	if ( cmdline_image.len ) {
+		if ( ( rc = register_image ( &cmdline_image ) ) != 0 ) {
+			DBGC ( colour, "RUNTIME could not register command "
+			       "line: %s\n", strerror ( rc ) );
+			goto err_register_image;
+		}
+	}
+
+	/* Drop our reference to the image */
+	image_put ( &cmdline_image );
+
+	return 0;
+
+ err_register_image:
+	image_put ( &cmdline_image );
+ err_alloc_cmdline_copy:
+	return rc;
+}
+
+/**
+ * Initialise initrd
+ *
+ * @ret rc		Return status code
+ */
+static int initrd_init ( void ) {
+	struct image *image;
+	int rc;
+
+	/* Do nothing if no initrd was specified */
+	if ( ! initrd_phys ) {
+		DBGC ( colour, "RUNTIME found no initrd\n" );
+		return 0;
+	}
+	if ( ! initrd_len ) {
+		DBGC ( colour, "RUNTIME found empty initrd\n" );
+		return 0;
+	}
+	DBGC ( colour, "RUNTIME found initrd at [%x,%x)\n",
+	       initrd_phys, ( initrd_phys + initrd_len ) );
+
+	/* Allocate image */
+	image = alloc_image ( NULL );
+	if ( ! image ) {
+		DBGC ( colour, "RUNTIME could not allocate image for "
+		       "initrd\n" );
+		rc = -ENOMEM;
+		goto err_alloc_image;
+	}
+	if ( ( rc = image_set_name ( image, "<INITRD>" ) ) != 0 ) {
+		DBGC ( colour, "RUNTIME could not set image name: %s\n",
+		       strerror ( rc ) );
+		goto err_set_name;
+	}
+
+	/* Allocate and copy initrd content */
+	image->data = umalloc ( initrd_len );
+	if ( ! image->data ) {
+		DBGC ( colour, "RUNTIME could not allocate %d bytes for "
+		       "initrd\n", initrd_len );
+		rc = -ENOMEM;
+		goto err_umalloc;
+	}
+	image->len = initrd_len;
+	memcpy_user ( image->data, 0, phys_to_user ( initrd_phys ), 0,
+		      initrd_len );
+
+	/* Mark initrd as consumed */
+	initrd_phys = 0;
+
+	/* Register image */
+	if ( ( rc = register_image ( image ) ) != 0 ) {
+		DBGC ( colour, "RUNTIME could not register initrd: %s\n",
+		       strerror ( rc ) );
+		goto err_register_image;
+	}
+
+	/* Drop our reference to the image */
+	image_put ( image );
+
+	return 0;
+
+ err_register_image:
+ err_umalloc:
+ err_set_name:
+	image_put ( image );
+ err_alloc_image:
+	return rc;
+}
+
+/**
+ * Initialise command line and initrd
+ *
+ */
+static void runtime_init ( void ) {
+	int rc;
+
+	/* Initialise command line */
+	if ( ( rc = cmdline_init() ) != 0 ) {
+		/* No way to report failure */
+		return;
+	}
+
+	/* Initialise initrd */
+	if ( ( rc = initrd_init() ) != 0 ) {
+		/* No way to report failure */
+		return;
+	}
+}
+
+/** Command line and initrd initialisation function */
+struct startup_fn runtime_startup_fn __startup_fn ( STARTUP_NORMAL ) = {
+	.name = "runtime",
+	.startup = runtime_init,
+};

src/arch/x86/core/stack.S

@@ -0,0 +1,21 @@
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL )
+
+	.arch i386
+
+#ifdef __x86_64__
+#define STACK_SIZE 8192
+#else
+#define STACK_SIZE 4096
+#endif
+
+/****************************************************************************
+ * Internal stack
+ ****************************************************************************
+ */
+	.section ".stack", "aw", @nobits
+	.align 8
+	.globl _stack
+_stack:
+	.space STACK_SIZE
+	.globl _estack
+_estack:

src/arch/x86/core/stack16.S

@@ -0,0 +1,15 @@
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL )
+
+	.arch i386
+
+/****************************************************************************
+ * Internal stack
+ ****************************************************************************
+ */
+	.section ".stack16", "aw", @nobits
+	.align 8
+	.globl _stack16
+_stack16:
+	.space 4096
+	.globl _estack16
+_estack16:

src/arch/x86/core/video_subr.c

@@ -0,0 +1,113 @@
+/*
+ *
+ * modified from linuxbios code
+ * by Cai Qiang <rimy2000@hotmail.com>
+ *
+ */
+
+#include "stddef.h"
+#include "string.h"
+#include <ipxe/io.h>
+#include <ipxe/console.h>
+#include <ipxe/init.h>
+#include "vga.h"
+#include <config/console.h>
+
+/* Set default console usage if applicable */
+#if ! ( defined ( CONSOLE_DIRECT_VGA ) && \
+	CONSOLE_EXPLICIT ( CONSOLE_DIRECT_VGA ) )
+#undef CONSOLE_DIRECT_VGA
+#define CONSOLE_DIRECT_VGA ( CONSOLE_USAGE_ALL & ~CONSOLE_USAGE_LOG )
+#endif
+
+struct console_driver vga_console __console_driver;
+
+static char *vidmem;		/* The video buffer */
+static int video_line, video_col;
+
+#define VIDBUFFER 0xB8000	
+
+static void memsetw(void *s, int c, unsigned int n)
+{
+	unsigned int i;
+	u16 *ss = (u16 *) s;
+
+	for (i = 0; i < n; i++) {
+		ss[i] = ( u16 ) c;
+	}
+}
+
+static void video_init(void)
+{
+	static int inited=0;
+
+	vidmem = (char *)phys_to_virt(VIDBUFFER);
+
+	if (!inited) {
+		video_line = 0;
+		video_col = 0;
+	
+	 	memsetw(vidmem, VGA_ATTR_CLR_WHT, 2*1024); //
+
+		inited=1;
+	}
+}
+
+static void video_scroll(void)
+{
+	int i;
+
+	memmove(vidmem, vidmem + COLS * 2, (LINES - 1) * COLS * 2);
+	for (i = (LINES - 1) * COLS * 2; i < LINES * COLS * 2; i += 2)
+		vidmem[i] = ' ';
+}
+
+static void vga_putc(int byte)
+{
+	if (byte == '\n') {
+		video_line++;
+		video_col = 0;
+
+	} else if (byte == '\r') {
+		video_col = 0;
+
+	} else if (byte == '\b') {
+		video_col--;
+
+	} else if (byte == '\t') {
+		video_col += 4;
+
+	} else if (byte == '\a') {
+		//beep
+		//beep(500);
+
+	} else {
+		vidmem[((video_col + (video_line *COLS)) * 2)] = byte;
+		vidmem[((video_col + (video_line *COLS)) * 2) +1] = VGA_ATTR_CLR_WHT;
+		video_col++;
+	}
+	if (video_col < 0) {
+		video_col = 0;
+	}
+	if (video_col >= COLS) {
+		video_line++;
+		video_col = 0;
+	}
+	if (video_line >= LINES) {
+		video_scroll();
+		video_line--;
+	}
+	// move the cursor
+	write_crtc((video_col + (video_line *COLS)) >> 8, CRTC_CURSOR_HI);
+	write_crtc((video_col + (video_line *COLS)) & 0x0ff, CRTC_CURSOR_LO);
+}
+
+struct console_driver vga_console __console_driver = {
+	.putchar = vga_putc,
+	.disabled = CONSOLE_DISABLED,
+	.usage = CONSOLE_DIRECT_VGA,
+};
+
+struct init_fn video_init_fn __init_fn ( INIT_EARLY ) = {
+	.initialise = video_init,
+};

src/arch/x86/core/vram_settings.c

@@ -0,0 +1,72 @@
+/*
+ * Copyright (C) 2015 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <ipxe/uaccess.h>
+#include <ipxe/settings.h>
+
+/** @file
+ *
+ * Video RAM dump
+ *
+ */
+
+/** Video RAM base address */
+#define VRAM_BASE 0xb8000
+
+/** Video RAM length */
+#define VRAM_LEN \
+	( 80 /* columns */ * 25 /* rows */ * 2 /* bytes per character */ )
+
+/**
+ * Fetch video RAM setting
+ *
+ * @v data		Buffer to fill with setting data
+ * @v len		Length of buffer
+ * @ret len		Length of setting data, or negative error
+ */
+static int vram_fetch ( void *data, size_t len ) {
+	userptr_t vram = phys_to_user ( VRAM_BASE );
+
+	/* Copy video RAM */
+	if ( len > VRAM_LEN )
+		len = VRAM_LEN;
+	copy_from_user ( data, vram, 0, len );
+
+	return VRAM_LEN;
+}
+
+/** Video RAM setting */
+const struct setting vram_setting __setting ( SETTING_MISC, vram ) = {
+	.name = "vram",
+	.description = "Video RAM",
+	.type = &setting_type_base64,
+	.scope = &builtin_scope,
+};
+
+/** Video RAM built-in setting */
+struct builtin_setting vram_builtin_setting __builtin_setting = {
+	.setting = &vram_setting,
+	.fetch = vram_fetch,
+};

src/arch/x86/core/x86_uart.c

@@ -0,0 +1,69 @@
+/*
+ * Copyright (C) 2014 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/** @file
+ *
+ * 16550-compatible UART
+ *
+ */
+
+#include <errno.h>
+#include <ipxe/uart.h>
+
+/** UART port bases */
+static uint16_t uart_base[] = {
+	[COM1] = 0x3f8,
+	[COM2] = 0x2f8,
+	[COM3] = 0x3e8,
+	[COM4] = 0x2e8,
+};
+
+/**
+ * Select UART port
+ *
+ * @v uart		UART
+ * @v port		Port number, or 0 to disable
+ * @ret rc		Return status code
+ */
+int uart_select ( struct uart *uart, unsigned int port ) {
+	int rc;
+
+	/* Set new UART base */
+	if ( port >= ( sizeof ( uart_base ) / sizeof ( uart_base[0] ) ) ) {
+		rc = -ENODEV;
+		goto err;
+	}
+	uart->base = ( ( void * ) ( intptr_t ) uart_base[port] );
+
+	/* Check that UART exists */
+	if ( ( rc = uart_exists ( uart ) ) != 0 )
+		goto err;
+
+	return 0;
+
+ err:
+	uart->base = NULL;
+	return rc;
+}

src/arch/x86/drivers/hyperv/hyperv.c

@@ -0,0 +1,820 @@
+/*
+ * Copyright (C) 2014 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/** @file
+ *
+ * Hyper-V driver
+ *
+ */
+
+#include <stdlib.h>
+#include <stdarg.h>
+#include <string.h>
+#include <unistd.h>
+#include <assert.h>
+#include <errno.h>
+#include <byteswap.h>
+#include <pic8259.h>
+#include <ipxe/malloc.h>
+#include <ipxe/device.h>
+#include <ipxe/timer.h>
+#include <ipxe/quiesce.h>
+#include <ipxe/cpuid.h>
+#include <ipxe/msr.h>
+#include <ipxe/hyperv.h>
+#include <ipxe/vmbus.h>
+#include "hyperv.h"
+
+/** Maximum time to wait for a message response
+ *
+ * This is a policy decision.
+ */
+#define HV_MESSAGE_MAX_WAIT_MS 1000
+
+/** Hyper-V timer frequency (fixed 10Mhz) */
+#define HV_TIMER_HZ 10000000
+
+/** Hyper-V timer scale factor (used to avoid 64-bit division) */
+#define HV_TIMER_SHIFT 18
+
+/**
+ * Convert a Hyper-V status code to an iPXE status code
+ *
+ * @v status		Hyper-V status code
+ * @ret rc		iPXE status code (before negation)
+ */
+#define EHV( status ) EPLATFORM ( EINFO_EPLATFORM, (status) )
+
+/**
+ * Allocate zeroed pages
+ *
+ * @v hv		Hyper-V hypervisor
+ * @v ...		Page addresses to fill in, terminated by NULL
+ * @ret rc		Return status code
+ */
+__attribute__ (( sentinel )) int
+hv_alloc_pages ( struct hv_hypervisor *hv, ... ) {
+	va_list args;
+	void **page;
+	int i;
+
+	/* Allocate and zero pages */
+	va_start ( args, hv );
+	for ( i = 0 ; ( ( page = va_arg ( args, void ** ) ) != NULL ); i++ ) {
+		*page = malloc_dma ( PAGE_SIZE, PAGE_SIZE );
+		if ( ! *page )
+			goto err_alloc;
+		memset ( *page, 0, PAGE_SIZE );
+	}
+	va_end ( args );
+
+	return 0;
+
+ err_alloc:
+	va_end ( args );
+	va_start ( args, hv );
+	for ( ; i >= 0 ; i-- ) {
+		page = va_arg ( args, void ** );
+		free_dma ( *page, PAGE_SIZE );
+	}
+	va_end ( args );
+	return -ENOMEM;
+}
+
+/**
+ * Free pages
+ *
+ * @v hv		Hyper-V hypervisor
+ * @v ...		Page addresses, terminated by NULL
+ */
+__attribute__ (( sentinel )) void
+hv_free_pages ( struct hv_hypervisor *hv, ... ) {
+	va_list args;
+	void *page;
+
+	va_start ( args, hv );
+	while ( ( page = va_arg ( args, void * ) ) != NULL )
+		free_dma ( page, PAGE_SIZE );
+	va_end ( args );
+}
+
+/**
+ * Allocate message buffer
+ *
+ * @v hv		Hyper-V hypervisor
+ * @ret rc		Return status code
+ */
+static int hv_alloc_message ( struct hv_hypervisor *hv ) {
+
+	/* Allocate buffer.  Must be aligned to at least 8 bytes and
+	 * must not cross a page boundary, so align on its own size.
+	 */
+	hv->message = malloc_dma ( sizeof ( *hv->message ),
+				   sizeof ( *hv->message ) );
+	if ( ! hv->message )
+		return -ENOMEM;
+
+	return 0;
+}
+
+/**
+ * Free message buffer
+ *
+ * @v hv		Hyper-V hypervisor
+ */
+static void hv_free_message ( struct hv_hypervisor *hv ) {
+
+	/* Free buffer */
+	free_dma ( hv->message, sizeof ( *hv->message ) );
+}
+
+/**
+ * Check whether or not we are running in Hyper-V
+ *
+ * @ret rc		Return status code
+ */
+static int hv_check_hv ( void ) {
+	struct x86_features features;
+	uint32_t interface_id;
+	uint32_t discard_ebx;
+	uint32_t discard_ecx;
+	uint32_t discard_edx;
+
+	/* Check for presence of a hypervisor (not necessarily Hyper-V) */
+	x86_features ( &features );
+	if ( ! ( features.intel.ecx & CPUID_FEATURES_INTEL_ECX_HYPERVISOR ) ) {
+		DBGC ( HV_INTERFACE_ID, "HV not running in a hypervisor\n" );
+		return -ENODEV;
+	}
+
+	/* Check that hypervisor is Hyper-V */
+	cpuid ( HV_CPUID_INTERFACE_ID, 0, &interface_id, &discard_ebx,
+		&discard_ecx, &discard_edx );
+	if ( interface_id != HV_INTERFACE_ID ) {
+		DBGC ( HV_INTERFACE_ID, "HV not running in Hyper-V (interface "
+		       "ID %#08x)\n", interface_id );
+		return -ENODEV;
+	}
+
+	return 0;
+}
+
+/**
+ * Check required features
+ *
+ * @v hv		Hyper-V hypervisor
+ * @ret rc		Return status code
+ */
+static int hv_check_features ( struct hv_hypervisor *hv ) {
+	uint32_t available;
+	uint32_t permissions;
+	uint32_t discard_ecx;
+	uint32_t discard_edx;
+
+	/* Check that required features and privileges are available */
+	cpuid ( HV_CPUID_FEATURES, 0, &available, &permissions, &discard_ecx,
+		&discard_edx );
+	if ( ! ( available & HV_FEATURES_AVAIL_HYPERCALL_MSR ) ) {
+		DBGC ( hv, "HV %p has no hypercall MSRs (features %08x:%08x)\n",
+		       hv, available, permissions );
+		return -ENODEV;
+	}
+	if ( ! ( available & HV_FEATURES_AVAIL_SYNIC_MSR ) ) {
+		DBGC ( hv, "HV %p has no SynIC MSRs (features %08x:%08x)\n",
+		       hv, available, permissions );
+		return -ENODEV;
+	}
+	if ( ! ( permissions & HV_FEATURES_PERM_POST_MESSAGES ) ) {
+		DBGC ( hv, "HV %p cannot post messages (features %08x:%08x)\n",
+		       hv, available, permissions );
+		return -EACCES;
+	}
+	if ( ! ( permissions & HV_FEATURES_PERM_SIGNAL_EVENTS ) ) {
+		DBGC ( hv, "HV %p cannot signal events (features %08x:%08x)",
+		       hv, available, permissions );
+		return -EACCES;
+	}
+
+	return 0;
+}
+
+/**
+ * Check that Gen 2 UEFI firmware is not running
+ *
+ * @v hv		Hyper-V hypervisor
+ * @ret rc		Return status code
+ *
+ * We must not steal ownership from the Gen 2 UEFI firmware, since
+ * doing so will cause an immediate crash.  Avoid this by checking for
+ * the guest OS identity known to be used by the Gen 2 UEFI firmware.
+ */
+static int hv_check_uefi ( struct hv_hypervisor *hv ) {
+	uint64_t guest_os_id;
+
+	/* Check for UEFI firmware's guest OS identity */
+	guest_os_id = rdmsr ( HV_X64_MSR_GUEST_OS_ID );
+	if ( guest_os_id == HV_GUEST_OS_ID_UEFI ) {
+		DBGC ( hv, "HV %p is owned by UEFI firmware\n", hv );
+		return -ENOTSUP;
+	}
+
+	return 0;
+}
+
+/**
+ * Map hypercall page
+ *
+ * @v hv		Hyper-V hypervisor
+ */
+static void hv_map_hypercall ( struct hv_hypervisor *hv ) {
+	union {
+		struct {
+			uint32_t ebx;
+			uint32_t ecx;
+			uint32_t edx;
+		} __attribute__ (( packed ));
+		char text[ 13 /* "bbbbccccdddd" + NUL */ ];
+	} vendor_id;
+	uint32_t build;
+	uint32_t version;
+	uint32_t discard_eax;
+	uint32_t discard_ecx;
+	uint32_t discard_edx;
+	uint64_t guest_os_id;
+	uint64_t hypercall;
+
+	/* Report guest OS identity */
+	guest_os_id = rdmsr ( HV_X64_MSR_GUEST_OS_ID );
+	if ( guest_os_id != 0 ) {
+		DBGC ( hv, "HV %p guest OS ID MSR was %#08llx\n",
+		       hv, guest_os_id );
+	}
+	guest_os_id = HV_GUEST_OS_ID_IPXE;
+	DBGC2 ( hv, "HV %p guest OS ID MSR is %#08llx\n", hv, guest_os_id );
+	wrmsr ( HV_X64_MSR_GUEST_OS_ID, guest_os_id );
+
+	/* Get hypervisor system identity (for debugging) */
+	cpuid ( HV_CPUID_VENDOR_ID, 0, &discard_eax, &vendor_id.ebx,
+		&vendor_id.ecx, &vendor_id.edx );
+	vendor_id.text[ sizeof ( vendor_id.text ) - 1 ] = '\0';
+	cpuid ( HV_CPUID_HYPERVISOR_ID, 0, &build, &version, &discard_ecx,
+		&discard_edx );
+	DBGC ( hv, "HV %p detected \"%s\" version %d.%d build %d\n", hv,
+	       vendor_id.text, ( version >> 16 ), ( version & 0xffff ), build );
+
+	/* Map hypercall page */
+	hypercall = rdmsr ( HV_X64_MSR_HYPERCALL );
+	hypercall &= ( PAGE_SIZE - 1 );
+	hypercall |= ( virt_to_phys ( hv->hypercall ) | HV_HYPERCALL_ENABLE );
+	DBGC2 ( hv, "HV %p hypercall MSR is %#08llx\n", hv, hypercall );
+	wrmsr ( HV_X64_MSR_HYPERCALL, hypercall );
+}
+
+/**
+ * Unmap hypercall page
+ *
+ * @v hv		Hyper-V hypervisor
+ */
+static void hv_unmap_hypercall ( struct hv_hypervisor *hv ) {
+	uint64_t hypercall;
+	uint64_t guest_os_id;
+
+	/* Unmap the hypercall page */
+	hypercall = rdmsr ( HV_X64_MSR_HYPERCALL );
+	hypercall &= ( ( PAGE_SIZE - 1 ) & ~HV_HYPERCALL_ENABLE );
+	DBGC2 ( hv, "HV %p hypercall MSR is %#08llx\n", hv, hypercall );
+	wrmsr ( HV_X64_MSR_HYPERCALL, hypercall );
+
+	/* Reset the guest OS identity */
+	guest_os_id = 0;
+	DBGC2 ( hv, "HV %p guest OS ID MSR is %#08llx\n", hv, guest_os_id );
+	wrmsr ( HV_X64_MSR_GUEST_OS_ID, guest_os_id );
+}
+
+/**
+ * Map synthetic interrupt controller
+ *
+ * @v hv		Hyper-V hypervisor
+ */
+static void hv_map_synic ( struct hv_hypervisor *hv ) {
+	uint64_t simp;
+	uint64_t siefp;
+	uint64_t scontrol;
+
+	/* Zero SynIC message and event pages */
+	memset ( hv->synic.message, 0, PAGE_SIZE );
+	memset ( hv->synic.event, 0, PAGE_SIZE );
+
+	/* Map SynIC message page */
+	simp = rdmsr ( HV_X64_MSR_SIMP );
+	simp &= ( PAGE_SIZE - 1 );
+	simp |= ( virt_to_phys ( hv->synic.message ) | HV_SIMP_ENABLE );
+	DBGC2 ( hv, "HV %p SIMP MSR is %#08llx\n", hv, simp );
+	wrmsr ( HV_X64_MSR_SIMP, simp );
+
+	/* Map SynIC event page */
+	siefp = rdmsr ( HV_X64_MSR_SIEFP );
+	siefp &= ( PAGE_SIZE - 1 );
+	siefp |= ( virt_to_phys ( hv->synic.event ) | HV_SIEFP_ENABLE );
+	DBGC2 ( hv, "HV %p SIEFP MSR is %#08llx\n", hv, siefp );
+	wrmsr ( HV_X64_MSR_SIEFP, siefp );
+
+	/* Enable SynIC */
+	scontrol = rdmsr ( HV_X64_MSR_SCONTROL );
+	scontrol |= HV_SCONTROL_ENABLE;
+	DBGC2 ( hv, "HV %p SCONTROL MSR is %#08llx\n", hv, scontrol );
+	wrmsr ( HV_X64_MSR_SCONTROL, scontrol );
+}
+
+/**
+ * Unmap synthetic interrupt controller, leaving SCONTROL untouched
+ *
+ * @v hv		Hyper-V hypervisor
+ */
+static void hv_unmap_synic_no_scontrol ( struct hv_hypervisor *hv ) {
+	uint64_t siefp;
+	uint64_t simp;
+
+	/* Unmap SynIC event page */
+	siefp = rdmsr ( HV_X64_MSR_SIEFP );
+	siefp &= ( ( PAGE_SIZE - 1 ) & ~HV_SIEFP_ENABLE );
+	DBGC2 ( hv, "HV %p SIEFP MSR is %#08llx\n", hv, siefp );
+	wrmsr ( HV_X64_MSR_SIEFP, siefp );
+
+	/* Unmap SynIC message page */
+	simp = rdmsr ( HV_X64_MSR_SIMP );
+	simp &= ( ( PAGE_SIZE - 1 ) & ~HV_SIMP_ENABLE );
+	DBGC2 ( hv, "HV %p SIMP MSR is %#08llx\n", hv, simp );
+	wrmsr ( HV_X64_MSR_SIMP, simp );
+}
+
+/**
+ * Unmap synthetic interrupt controller
+ *
+ * @v hv		Hyper-V hypervisor
+ */
+static void hv_unmap_synic ( struct hv_hypervisor *hv ) {
+	uint64_t scontrol;
+
+	/* Disable SynIC */
+	scontrol = rdmsr ( HV_X64_MSR_SCONTROL );
+	scontrol &= ~HV_SCONTROL_ENABLE;
+	DBGC2 ( hv, "HV %p SCONTROL MSR is %#08llx\n", hv, scontrol );
+	wrmsr ( HV_X64_MSR_SCONTROL, scontrol );
+
+	/* Unmap SynIC event and message pages */
+	hv_unmap_synic_no_scontrol ( hv );
+}
+
+/**
+ * Enable synthetic interrupt
+ *
+ * @v hv		Hyper-V hypervisor
+ * @v sintx		Synthetic interrupt number
+ */
+void hv_enable_sint ( struct hv_hypervisor *hv, unsigned int sintx ) {
+	unsigned long msr = HV_X64_MSR_SINT ( sintx );
+	uint64_t sint;
+
+	/* Enable synthetic interrupt
+	 *
+	 * We have to enable the interrupt, otherwise messages will
+	 * not be delivered (even though the documentation implies
+	 * that polling for messages is possible).  We enable AutoEOI
+	 * and hook the interrupt to the obsolete IRQ13 (FPU
+	 * exception) vector, which will be implemented as a no-op.
+	 */
+	sint = rdmsr ( msr );
+	sint &= ~( HV_SINT_MASKED | HV_SINT_VECTOR_MASK );
+	sint |= ( HV_SINT_AUTO_EOI |
+		  HV_SINT_VECTOR ( IRQ_INT ( 13 /* See comment above */ ) ) );
+	DBGC2 ( hv, "HV %p SINT%d MSR is %#08llx\n", hv, sintx, sint );
+	wrmsr ( msr, sint );
+}
+
+/**
+ * Disable synthetic interrupt
+ *
+ * @v hv		Hyper-V hypervisor
+ * @v sintx		Synthetic interrupt number
+ */
+void hv_disable_sint ( struct hv_hypervisor *hv, unsigned int sintx ) {
+	unsigned long msr = HV_X64_MSR_SINT ( sintx );
+	uint64_t sint;
+
+	/* Do nothing if interrupt is already disabled */
+	sint = rdmsr ( msr );
+	if ( sint & HV_SINT_MASKED )
+		return;
+
+	/* Disable synthetic interrupt */
+	sint &= ~HV_SINT_AUTO_EOI;
+	sint |= HV_SINT_MASKED;
+	DBGC2 ( hv, "HV %p SINT%d MSR is %#08llx\n", hv, sintx, sint );
+	wrmsr ( msr, sint );
+}
+
+/**
+ * Post message
+ *
+ * @v hv		Hyper-V hypervisor
+ * @v id		Connection ID
+ * @v type		Message type
+ * @v data		Message
+ * @v len		Length of message
+ * @ret rc		Return status code
+ */
+int hv_post_message ( struct hv_hypervisor *hv, unsigned int id,
+		      unsigned int type, const void *data, size_t len ) {
+	struct hv_post_message *msg = &hv->message->posted;
+	int status;
+	int rc;
+
+	/* Sanity check */
+	assert ( len <= sizeof ( msg->data ) );
+
+	/* Construct message */
+	memset ( msg, 0, sizeof ( *msg ) );
+	msg->id = cpu_to_le32 ( id );
+	msg->type = cpu_to_le32 ( type );
+	msg->len = cpu_to_le32 ( len );
+	memcpy ( msg->data, data, len );
+	DBGC2 ( hv, "HV %p connection %d posting message type %#08x:\n",
+		hv, id, type );
+	DBGC2_HDA ( hv, 0, msg->data, len );
+
+	/* Post message */
+	if ( ( status = hv_call ( hv, HV_POST_MESSAGE, msg, NULL ) ) != 0 ) {
+		rc = -EHV ( status );
+		DBGC ( hv, "HV %p could not post message to %#08x: %s\n",
+		       hv, id, strerror ( rc ) );
+		return rc;
+	}
+
+	return 0;
+}
+
+/**
+ * Wait for received message
+ *
+ * @v hv		Hyper-V hypervisor
+ * @v sintx		Synthetic interrupt number
+ * @ret rc		Return status code
+ */
+int hv_wait_for_message ( struct hv_hypervisor *hv, unsigned int sintx ) {
+	struct hv_message *msg = &hv->message->received;
+	struct hv_message *src = &hv->synic.message[sintx];
+	unsigned int retries;
+	size_t len;
+
+	/* Wait for message to arrive */
+	for ( retries = 0 ; retries < HV_MESSAGE_MAX_WAIT_MS ; retries++ ) {
+
+		/* Check for message */
+		if ( src->type ) {
+
+			/* Copy message */
+			memset ( msg, 0, sizeof ( *msg ) );
+			len = src->len;
+			assert ( len <= sizeof ( *msg ) );
+			memcpy ( msg, src,
+				 ( offsetof ( typeof ( *msg ), data ) + len ) );
+			DBGC2 ( hv, "HV %p SINT%d received message type "
+				"%#08x:\n", hv, sintx,
+				le32_to_cpu ( msg->type ) );
+			DBGC2_HDA ( hv, 0, msg->data, len );
+
+			/* Consume message */
+			src->type = 0;
+
+			return 0;
+		}
+
+		/* Trigger message delivery */
+		wrmsr ( HV_X64_MSR_EOM, 0 );
+
+		/* Delay */
+		mdelay ( 1 );
+	}
+
+	DBGC ( hv, "HV %p SINT%d timed out waiting for message\n",
+	       hv, sintx );
+	return -ETIMEDOUT;
+}
+
+/**
+ * Signal event
+ *
+ * @v hv		Hyper-V hypervisor
+ * @v id		Connection ID
+ * @v flag		Flag number
+ * @ret rc		Return status code
+ */
+int hv_signal_event ( struct hv_hypervisor *hv, unsigned int id,
+		      unsigned int flag ) {
+	struct hv_signal_event *event = &hv->message->signalled;
+	int status;
+	int rc;
+
+	/* Construct event */
+	memset ( event, 0, sizeof ( *event ) );
+	event->id = cpu_to_le32 ( id );
+	event->flag = cpu_to_le16 ( flag );
+
+	/* Signal event */
+	if ( ( status = hv_call ( hv, HV_SIGNAL_EVENT, event, NULL ) ) != 0 ) {
+		rc = -EHV ( status );
+		DBGC ( hv, "HV %p could not signal event to %#08x: %s\n",
+		       hv, id, strerror ( rc ) );
+		return rc;
+	}
+
+	return 0;
+}
+
+/**
+ * Probe root device
+ *
+ * @v rootdev		Root device
+ * @ret rc		Return status code
+ */
+static int hv_probe ( struct root_device *rootdev ) {
+	struct hv_hypervisor *hv;
+	int rc;
+
+	/* Check we are running in Hyper-V */
+	if ( ( rc = hv_check_hv() ) != 0 )
+		goto err_check_hv;
+
+	/* Allocate and initialise structure */
+	hv = zalloc ( sizeof ( *hv ) );
+	if ( ! hv ) {
+		rc = -ENOMEM;
+		goto err_alloc;
+	}
+
+	/* Check features */
+	if ( ( rc = hv_check_features ( hv ) ) != 0 )
+		goto err_check_features;
+
+	/* Check that Gen 2 UEFI firmware is not running */
+	if ( ( rc = hv_check_uefi ( hv ) ) != 0 )
+		goto err_check_uefi;
+
+	/* Allocate pages */
+	if ( ( rc = hv_alloc_pages ( hv, &hv->hypercall, &hv->synic.message,
+				     &hv->synic.event, NULL ) ) != 0 )
+		goto err_alloc_pages;
+
+	/* Allocate message buffer */
+	if ( ( rc = hv_alloc_message ( hv ) ) != 0 )
+		goto err_alloc_message;
+
+	/* Map hypercall page */
+	hv_map_hypercall ( hv );
+
+	/* Map synthetic interrupt controller */
+	hv_map_synic ( hv );
+
+	/* Probe Hyper-V devices */
+	if ( ( rc = vmbus_probe ( hv, &rootdev->dev ) ) != 0 )
+		goto err_vmbus_probe;
+
+	rootdev_set_drvdata ( rootdev, hv );
+	return 0;
+
+	vmbus_remove ( hv, &rootdev->dev );
+ err_vmbus_probe:
+	hv_unmap_synic ( hv );
+	hv_unmap_hypercall ( hv );
+	hv_free_message ( hv );
+ err_alloc_message:
+	hv_free_pages ( hv, hv->hypercall, hv->synic.message, hv->synic.event,
+			NULL );
+ err_alloc_pages:
+ err_check_uefi:
+ err_check_features:
+	free ( hv );
+ err_alloc:
+ err_check_hv:
+	return rc;
+}
+
+/**
+ * Remove root device
+ *
+ * @v rootdev		Root device
+ */
+static void hv_remove ( struct root_device *rootdev ) {
+	struct hv_hypervisor *hv = rootdev_get_drvdata ( rootdev );
+
+	vmbus_remove ( hv, &rootdev->dev );
+	hv_unmap_synic ( hv );
+	hv_unmap_hypercall ( hv );
+	hv_free_message ( hv );
+	hv_free_pages ( hv, hv->hypercall, hv->synic.message, hv->synic.event,
+			NULL );
+	free ( hv );
+	rootdev_set_drvdata ( rootdev, NULL );
+}
+
+/** Hyper-V root device driver */
+static struct root_driver hv_root_driver = {
+	.probe = hv_probe,
+	.remove = hv_remove,
+};
+
+/** Hyper-V root device */
+struct root_device hv_root_device __root_device = {
+	.dev = { .name = "Hyper-V" },
+	.driver = &hv_root_driver,
+};
+
+/**
+ * Quiesce system
+ *
+ */
+static void hv_quiesce ( void ) {
+	struct hv_hypervisor *hv = rootdev_get_drvdata ( &hv_root_device );
+	unsigned int i;
+
+	/* Do nothing if we are not running in Hyper-V */
+	if ( ! hv )
+		return;
+
+	/* The "enlightened" portions of the Windows Server 2016 boot
+	 * process will not cleanly take ownership of an active
+	 * Hyper-V connection.  Experimentation shows that the minimum
+	 * requirement is that we disable the SynIC message page
+	 * (i.e. zero the SIMP MSR).
+	 *
+	 * We cannot perform a full shutdown of the Hyper-V
+	 * connection.  Experimentation shows that if we disable the
+	 * SynIC (i.e. zero the SCONTROL MSR) then Windows Server 2016
+	 * will enter an indefinite wait loop.
+	 *
+	 * Attempt to create a safe handover environment by resetting
+	 * all MSRs except for SCONTROL.
+	 *
+	 * Note that we do not shut down our VMBus devices, since we
+	 * may need to unquiesce the system and continue operation.
+	 */
+
+	/* Disable all synthetic interrupts */
+	for ( i = 0 ; i <= HV_SINT_MAX ; i++ )
+		hv_disable_sint ( hv, i );
+
+	/* Unmap synthetic interrupt controller, leaving SCONTROL
+	 * enabled (see above).
+	 */
+	hv_unmap_synic_no_scontrol ( hv );
+
+	/* Unmap hypercall page */
+	hv_unmap_hypercall ( hv );
+
+	DBGC ( hv, "HV %p quiesced\n", hv );
+}
+
+/**
+ * Unquiesce system
+ *
+ */
+static void hv_unquiesce ( void ) {
+	struct hv_hypervisor *hv = rootdev_get_drvdata ( &hv_root_device );
+	uint64_t simp;
+	int rc;
+
+	/* Do nothing if we are not running in Hyper-V */
+	if ( ! hv )
+		return;
+
+	/* Experimentation shows that the "enlightened" portions of
+	 * Windows Server 2016 will break our Hyper-V connection at
+	 * some point during a SAN boot.  Surprisingly it does not
+	 * change the guest OS ID MSR, but it does leave the SynIC
+	 * message page disabled.
+	 *
+	 * Our own explicit quiescing procedure will also disable the
+	 * SynIC message page.  We can therefore use the SynIC message
+	 * page enable bit as a heuristic to determine when we need to
+	 * reestablish our Hyper-V connection.
+	 */
+	simp = rdmsr ( HV_X64_MSR_SIMP );
+	if ( simp & HV_SIMP_ENABLE )
+		return;
+
+	/* Remap hypercall page */
+	hv_map_hypercall ( hv );
+
+	/* Remap synthetic interrupt controller */
+	hv_map_synic ( hv );
+
+	/* Reset Hyper-V devices */
+	if ( ( rc = vmbus_reset ( hv, &hv_root_device.dev ) ) != 0 ) {
+		DBGC ( hv, "HV %p could not unquiesce: %s\n",
+		       hv, strerror ( rc ) );
+		/* Nothing we can do */
+		return;
+	}
+}
+
+/** Hyper-V quiescer */
+struct quiescer hv_quiescer __quiescer = {
+	.quiesce = hv_quiesce,
+	.unquiesce = hv_unquiesce,
+};
+
+/**
+ * Probe timer
+ *
+ * @ret rc		Return status code
+ */
+static int hv_timer_probe ( void ) {
+	uint32_t available;
+	uint32_t discard_ebx;
+	uint32_t discard_ecx;
+	uint32_t discard_edx;
+	int rc;
+
+	/* Check we are running in Hyper-V */
+	if ( ( rc = hv_check_hv() ) != 0 )
+		return rc;
+
+	/* Check for available reference counter */
+	cpuid ( HV_CPUID_FEATURES, 0, &available, &discard_ebx, &discard_ecx,
+		&discard_edx );
+	if ( ! ( available & HV_FEATURES_AVAIL_TIME_REF_COUNT_MSR ) ) {
+		DBGC ( HV_INTERFACE_ID, "HV has no time reference counter\n" );
+		return -ENODEV;
+	}
+
+	return 0;
+}
+
+/**
+ * Get current system time in ticks
+ *
+ * @ret ticks		Current time, in ticks
+ */
+static unsigned long hv_currticks ( void ) {
+
+	/* Calculate time using a combination of bit shifts and
+	 * multiplication (to avoid a 64-bit division).
+	 */
+	return ( ( rdmsr ( HV_X64_MSR_TIME_REF_COUNT ) >> HV_TIMER_SHIFT ) *
+		 ( TICKS_PER_SEC / ( HV_TIMER_HZ >> HV_TIMER_SHIFT ) ) );
+}
+
+/**
+ * Delay for a fixed number of microseconds
+ *
+ * @v usecs		Number of microseconds for which to delay
+ */
+static void hv_udelay ( unsigned long usecs ) {
+	uint32_t start;
+	uint32_t elapsed;
+	uint32_t threshold;
+
+	/* Spin until specified number of 10MHz ticks have elapsed */
+	start = rdmsr ( HV_X64_MSR_TIME_REF_COUNT );
+	threshold = ( usecs * ( HV_TIMER_HZ / 1000000 ) );
+	do {
+		elapsed = ( rdmsr ( HV_X64_MSR_TIME_REF_COUNT ) - start );
+	} while ( elapsed < threshold );
+}
+
+/** Hyper-V timer */
+struct timer hv_timer __timer ( TIMER_PREFERRED ) = {
+	.name = "Hyper-V",
+	.probe = hv_timer_probe,
+	.currticks = hv_currticks,
+	.udelay = hv_udelay,
+};
+
+/* Drag in objects via hv_root_device */
+REQUIRING_SYMBOL ( hv_root_device );
+
+/* Drag in netvsc driver */
+REQUIRE_OBJECT ( netvsc );

src/arch/x86/drivers/hyperv/hyperv.h

@@ -0,0 +1,63 @@
+#ifndef _HYPERV_H
+#define _HYPERV_H
+
+/** @file
+ *
+ * Hyper-V driver
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/** Get vendor identification */
+#define HV_CPUID_VENDOR_ID 0x40000000UL
+
+/** Get interface identification */
+#define HV_CPUID_INTERFACE_ID 0x40000001UL
+
+/** Get hypervisor identification */
+#define HV_CPUID_HYPERVISOR_ID 0x40000002UL
+
+/** Get hypervisor features */
+#define HV_CPUID_FEATURES 0x40000003UL
+
+/** Time reference counter MSR is available */
+#define HV_FEATURES_AVAIL_TIME_REF_COUNT_MSR 0x00000002UL
+
+/** SynIC MSRs are available */
+#define HV_FEATURES_AVAIL_SYNIC_MSR 0x00000004UL
+
+/** Hypercall MSRs are available */
+#define HV_FEATURES_AVAIL_HYPERCALL_MSR 0x00000020UL
+
+/** Guest may post messages */
+#define HV_FEATURES_PERM_POST_MESSAGES 0x00000010UL
+
+/** Guest may signal events */
+#define HV_FEATURES_PERM_SIGNAL_EVENTS 0x00000020UL
+
+/** Guest OS identity MSR */
+#define HV_X64_MSR_GUEST_OS_ID 0x40000000UL
+
+/** Hypercall page MSR */
+#define HV_X64_MSR_HYPERCALL 0x40000001UL
+
+/** Time reference MSR */
+#define HV_X64_MSR_TIME_REF_COUNT 0x40000020UL
+
+/** SynIC control MSR */
+#define HV_X64_MSR_SCONTROL 0x40000080UL
+
+/** SynIC event flags page MSR */
+#define HV_X64_MSR_SIEFP 0x40000082UL
+
+/** SynIC message page MSR */
+#define HV_X64_MSR_SIMP 0x40000083UL
+
+/** SynIC end of message MSR */
+#define HV_X64_MSR_EOM 0x40000084UL
+
+/** SynIC interrupt source MSRs */
+#define HV_X64_MSR_SINT(x) ( 0x40000090UL + (x) )
+
+#endif /* _HYPERV_H */

src/arch/x86/drivers/net/undi.c

@@ -0,0 +1,136 @@
+/*
+ * Copyright (C) 2007 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER );
+
+#include <stdint.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <ipxe/pci.h>
+#include <undi.h>
+#include <undirom.h>
+#include <undiload.h>
+#include <undinet.h>
+#include <undipreload.h>
+
+/** @file
+ *
+ * UNDI PCI driver
+ *
+ */
+
+/**
+ * Find UNDI ROM for PCI device
+ *
+ * @v pci		PCI device
+ * @ret undirom		UNDI ROM, or NULL
+ *
+ * Try to find a driver for this device.  Try an exact match on the
+ * ROM address first, then fall back to a vendor/device ID match only
+ */
+static struct undi_rom * undipci_find_rom ( struct pci_device *pci ) {
+	struct undi_rom *undirom;
+	unsigned long rombase;
+	
+	rombase = pci_bar_start ( pci, PCI_ROM_ADDRESS );
+	undirom = undirom_find_pci ( pci->vendor, pci->device, rombase );
+	if ( ! undirom )
+		undirom = undirom_find_pci ( pci->vendor, pci->device, 0 );
+	return undirom;
+}
+
+/**
+ * Probe PCI device
+ *
+ * @v pci		PCI device
+ * @v id		PCI ID
+ * @ret rc		Return status code
+ */
+static int undipci_probe ( struct pci_device *pci ) {
+	struct undi_device *undi;
+	struct undi_rom *undirom;
+	int rc;
+
+	/* Allocate UNDI device structure */
+	undi = zalloc ( sizeof ( *undi ) );
+	if ( ! undi )
+		return -ENOMEM;
+	pci_set_drvdata ( pci, undi );
+
+	/* Find/create our pixie */
+	if ( preloaded_undi.pci_busdevfn == pci->busdevfn ) {
+		/* Claim preloaded UNDI device */
+		DBGC ( undi, "UNDI %p using preloaded UNDI device\n", undi );
+		memcpy ( undi, &preloaded_undi, sizeof ( *undi ) );
+		memset ( &preloaded_undi, 0, sizeof ( preloaded_undi ) );
+	} else {
+		/* Find UNDI ROM for PCI device */
+		if ( ! ( undirom = undipci_find_rom ( pci ) ) ) {
+			rc = -ENODEV;
+			goto err_find_rom;
+		}
+
+		/* Call UNDI ROM loader to create pixie */
+		if ( ( rc = undi_load_pci ( undi, undirom,
+					    pci->busdevfn ) ) != 0 ) {
+			goto err_load_pci;
+		}
+	}
+
+	/* Create network device */
+	if ( ( rc = undinet_probe ( undi, &pci->dev ) ) != 0 )
+		goto err_undinet_probe;
+	
+	return 0;
+
+ err_undinet_probe:
+	undi_unload ( undi );
+ err_find_rom:
+ err_load_pci:
+	free ( undi );
+	pci_set_drvdata ( pci, NULL );
+	return rc;
+}
+
+/**
+ * Remove PCI device
+ *
+ * @v pci	PCI device
+ */
+static void undipci_remove ( struct pci_device *pci ) {
+	struct undi_device *undi = pci_get_drvdata ( pci );
+
+	undinet_remove ( undi );
+	undi_unload ( undi );
+	free ( undi );
+	pci_set_drvdata ( pci, NULL );
+}
+
+static struct pci_device_id undipci_nics[] = {
+	PCI_ROM ( 0xffff, 0xffff, "undipci", "UNDI (PCI)", 0 ),
+};
+
+struct pci_driver undipci_driver __pci_driver_fallback = {
+	.ids = undipci_nics,
+	.id_count = ( sizeof ( undipci_nics ) / sizeof ( undipci_nics[0] ) ),
+	.class = PCI_CLASS_ID ( PCI_CLASS_NETWORK, PCI_ANY_ID, PCI_ANY_ID ),
+	.probe = undipci_probe,
+	.remove = undipci_remove,
+};

src/arch/x86/drivers/net/undiisr.S

@@ -0,0 +1,87 @@
+FILE_LICENCE ( GPL2_OR_LATER )
+
+#define PXENV_UNDI_ISR 0x0014
+#define PXENV_UNDI_ISR_IN_START 1
+#define PXENV_UNDI_ISR_OUT_OURS 0
+#define PXENV_UNDI_ISR_OUT_NOT_OURS 1
+
+#define IRQ_PIC_CUTOFF 8
+#define ICR_EOI_NON_SPECIFIC 0x20
+#define PIC1_ICR 0x20
+#define PIC2_ICR 0xa0
+	
+	.text
+	.arch i386
+	.code16
+
+	.section ".text16", "ax", @progbits
+	.globl undiisr
+undiisr:
+	
+	/* Preserve registers */
+	pushw	%ds
+	pushw	%es
+	pushw	%fs
+	pushw	%gs
+	pushfl
+	pushal
+
+	/* Set up our segment registers */
+	movw	%cs:rm_ds, %ax
+	movw	%ax, %ds
+
+	/* Check that we have an UNDI entry point */
+	cmpw	$0, undinet_entry_point
+	je	chain
+	
+	/* Issue UNDI API call */
+	movw	%ax, %es
+	movw	$undinet_params, %di
+	movw	$PXENV_UNDI_ISR, %bx
+	movw	$PXENV_UNDI_ISR_IN_START, funcflag
+	pushw	%es
+	pushw	%di
+	pushw	%bx
+	lcall	*undinet_entry_point
+	cli	/* Just in case */
+	addw	$6, %sp
+	cmpw	$PXENV_UNDI_ISR_OUT_OURS, funcflag
+	jne	eoi
+	
+trig:	/* Record interrupt occurence */
+	incb	undiisr_trigger_count
+
+eoi:	/* Send EOI */
+	movb	$ICR_EOI_NON_SPECIFIC, %al
+	cmpb	$IRQ_PIC_CUTOFF, undiisr_irq
+	jb	1f
+	outb	%al, $PIC2_ICR
+1:	outb	%al, $PIC1_ICR
+	jmp	exit
+	
+chain:	/* Chain to next handler */
+	pushfw
+	lcall	*undiisr_next_handler
+	
+exit:	/* Restore registers and return */
+	cli
+	popal
+	movzwl	%sp, %esp
+	addr32	movl -20(%esp), %esp	/* %esp isn't restored by popal */
+	popfl
+	popw	%gs
+	popw	%fs
+	popw	%es
+	popw	%ds
+	iret
+
+	.section ".data16", "aw", @progbits
+undinet_params:
+status:			.word	0
+funcflag:		.word	0
+bufferlength:		.word	0
+framelength:		.word	0
+frameheaderlength:	.word	0
+frame:			.word	0, 0
+prottype:		.byte	0
+pkttype:		.byte	0

src/arch/x86/drivers/net/undiload.c

@@ -0,0 +1,186 @@
+/*
+ * Copyright (C) 2007 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+#include <pxe.h>
+#include <realmode.h>
+#include <bios.h>
+#include <pnpbios.h>
+#include <basemem.h>
+#include <ipxe/pci.h>
+#include <undi.h>
+#include <undirom.h>
+#include <undiload.h>
+
+/** @file
+ *
+ * UNDI load/unload
+ *
+ */
+
+/* Disambiguate the various error causes */
+#define EINFO_EUNDILOAD							\
+	__einfo_uniqify ( EINFO_EPLATFORM, 0x01,			\
+			  "UNDI loader error" )
+#define EUNDILOAD( status ) EPLATFORM ( EINFO_EUNDILOAD, status )
+
+/** Parameter block for calling UNDI loader */
+static struct s_UNDI_LOADER __bss16 ( undi_loader );
+#define undi_loader __use_data16 ( undi_loader )
+
+/** UNDI loader entry point */
+static SEGOFF16_t __bss16 ( undi_loader_entry );
+#define undi_loader_entry __use_data16 ( undi_loader_entry )
+
+/**
+ * Call UNDI loader to create a pixie
+ *
+ * @v undi		UNDI device
+ * @v undirom		UNDI ROM
+ * @ret rc		Return status code
+ */
+int undi_load ( struct undi_device *undi, struct undi_rom *undirom ) {
+	struct s_PXE ppxe;
+	unsigned int fbms_seg;
+	uint16_t exit;
+	int rc;
+
+	/* Only one UNDI instance may be loaded at any given time */
+	if ( undi_loader_entry.segment ) {
+		DBG ( "UNDI %p cannot load multiple instances\n", undi );
+		rc = -EBUSY;
+		goto err_multiple;
+	}
+
+	/* Set up START_UNDI parameters */
+	memset ( &undi_loader, 0, sizeof ( undi_loader ) );
+	undi_loader.AX = undi->pci_busdevfn;
+	undi_loader.BX = undi->isapnp_csn;
+	undi_loader.DX = undi->isapnp_read_port;
+	undi_loader.ES = BIOS_SEG;
+	undi_loader.DI = find_pnp_bios();
+
+	/* Allocate base memory for PXE stack */
+	undi->restore_fbms = get_fbms();
+	fbms_seg = ( undi->restore_fbms << 6 );
+	fbms_seg -= ( ( undirom->code_size + 0x0f ) >> 4 );
+	undi_loader.UNDI_CS = fbms_seg;
+	fbms_seg -= ( ( undirom->data_size + 0x0f ) >> 4 );
+	undi_loader.UNDI_DS = fbms_seg;
+	undi->fbms = ( fbms_seg >> 6 );
+	set_fbms ( undi->fbms );
+	DBGC ( undi, "UNDI %p allocated [%d,%d) kB of base memory\n",
+	       undi, undi->fbms, undi->restore_fbms );
+
+	/* Debug info */
+	DBGC ( undi, "UNDI %p loading ROM %p to CS %04x:%04zx DS %04x:%04zx "
+	       "for ", undi, undirom, undi_loader.UNDI_CS, undirom->code_size,
+	       undi_loader.UNDI_DS, undirom->data_size );
+	if ( undi->pci_busdevfn != UNDI_NO_PCI_BUSDEVFN ) {
+		unsigned int bus = ( undi->pci_busdevfn >> 8 );
+		unsigned int devfn = ( undi->pci_busdevfn & 0xff );
+		DBGC ( undi, "PCI %02x:%02x.%x\n",
+		       bus, PCI_SLOT ( devfn ), PCI_FUNC ( devfn ) );
+	}
+	if ( undi->isapnp_csn != UNDI_NO_ISAPNP_CSN ) {
+		DBGC ( undi, "ISAPnP(%04x) CSN %04x\n",
+		       undi->isapnp_read_port, undi->isapnp_csn );
+	}
+
+	/* Call loader */
+	undi_loader_entry = undirom->loader_entry;
+	__asm__ __volatile__ ( REAL_CODE ( "pushl %%ebp\n\t" /* gcc bug */
+					   "pushw %%ds\n\t"
+					   "pushw %%ax\n\t"
+					   "lcall *undi_loader_entry\n\t"
+					   "popl %%ebp\n\t" /* discard */
+					   "popl %%ebp\n\t" /* gcc bug */ )
+			       : "=a" ( exit )
+			       : "a" ( __from_data16 ( &undi_loader ) )
+			       : "ebx", "ecx", "edx", "esi", "edi" );
+	if ( exit != PXENV_EXIT_SUCCESS ) {
+		rc = -EUNDILOAD ( undi_loader.Status );
+		DBGC ( undi, "UNDI %p loader failed: %s\n",
+		       undi, strerror ( rc ) );
+		goto err_loader;
+	}
+
+	/* Populate PXE device structure */
+	undi->pxenv = undi_loader.PXENVptr;
+	undi->ppxe = undi_loader.PXEptr;
+	copy_from_real ( &ppxe, undi->ppxe.segment, undi->ppxe.offset,
+			 sizeof ( ppxe ) );
+	undi->entry = ppxe.EntryPointSP;
+	DBGC ( undi, "UNDI %p loaded PXENV+ %04x:%04x !PXE %04x:%04x "
+	       "entry %04x:%04x\n", undi, undi->pxenv.segment,
+	       undi->pxenv.offset, undi->ppxe.segment, undi->ppxe.offset,
+	       undi->entry.segment, undi->entry.offset );
+
+	return 0;
+
+ err_loader:
+	set_fbms ( undi->restore_fbms );
+	memset ( &undi_loader_entry, 0, sizeof ( undi_loader_entry ) );
+ err_multiple:
+	return rc;
+}
+
+/**
+ * Unload a pixie
+ *
+ * @v undi		UNDI device
+ * @ret rc		Return status code
+ *
+ * Erases the PXENV+ and !PXE signatures, and frees the used base
+ * memory (if possible).
+ */
+int undi_unload ( struct undi_device *undi ) {
+	static uint32_t dead = 0xdeaddead;
+
+	DBGC ( undi, "UNDI %p unloading\n", undi );
+
+	/* Clear entry point */
+	memset ( &undi_loader_entry, 0, sizeof ( undi_loader_entry ) );
+
+	/* Erase signatures */
+	if ( undi->pxenv.segment )
+		put_real ( dead, undi->pxenv.segment, undi->pxenv.offset );
+	if ( undi->ppxe.segment )
+		put_real ( dead, undi->ppxe.segment, undi->ppxe.offset );
+
+	/* Free base memory, if possible */
+	if ( undi->fbms == get_fbms() ) {
+		DBGC ( undi, "UNDI %p freeing [%d,%d) kB of base memory\n",
+		       undi, undi->fbms, undi->restore_fbms );
+		set_fbms ( undi->restore_fbms );
+		return 0;
+	} else {
+		DBGC ( undi, "UNDI %p leaking [%d,%d) kB of base memory\n",
+		       undi, undi->fbms, undi->restore_fbms );
+		return -EBUSY;
+	}
+}

src/arch/x86/drivers/net/undionly.c

@@ -0,0 +1,146 @@
+/*
+ * Copyright (C) 2007 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <ipxe/device.h>
+#include <ipxe/init.h>
+#include <ipxe/pci.h>
+#include <undi.h>
+#include <undinet.h>
+#include <undipreload.h>
+
+/** @file
+ *
+ * "Pure" UNDI driver
+ *
+ * This is the UNDI driver without explicit support for PCI or any
+ * other bus type.  It is capable only of using the preloaded UNDI
+ * device.  It must not be combined in an image with any other
+ * drivers.
+ *
+ * If you want a PXE-loadable image that contains only the UNDI
+ * driver, build "bin/undionly.kpxe".
+ *
+ * If you want any other image format, or any other drivers in
+ * addition to the UNDI driver, build e.g. "bin/undi.dsk".
+ */
+
+/** UNDI root bus device */
+static struct device undibus_dev;
+
+/**
+ * Probe UNDI root bus
+ *
+ * @v rootdev		UNDI bus root device
+ *
+ * Scans the UNDI bus for devices and registers all devices it can
+ * find.
+ */
+static int undibus_probe ( struct root_device *rootdev ) {
+	struct undi_device *undi = &preloaded_undi;
+	struct device *dev = &undibus_dev;
+	int rc;
+
+	/* Check for a valie preloaded UNDI device */
+	if ( ! undi->entry.segment ) {
+		DBG ( "No preloaded UNDI device found!\n" );
+		return -ENODEV;
+	}
+
+	/* Add to device hierarchy */
+	dev->driver_name = "undionly";
+	if ( undi->pci_busdevfn != UNDI_NO_PCI_BUSDEVFN ) {
+		dev->desc.bus_type = BUS_TYPE_PCI;
+		dev->desc.location = undi->pci_busdevfn;
+		dev->desc.vendor = undi->pci_vendor;
+		dev->desc.device = undi->pci_device;
+		snprintf ( dev->name, sizeof ( dev->name ),
+			   "0000:%02x:%02x.%x", PCI_BUS ( undi->pci_busdevfn ),
+			   PCI_SLOT ( undi->pci_busdevfn ),
+			   PCI_FUNC ( undi->pci_busdevfn ) );
+	} else if ( undi->isapnp_csn != UNDI_NO_ISAPNP_CSN ) {
+		dev->desc.bus_type = BUS_TYPE_ISAPNP;
+		snprintf ( dev->name, sizeof ( dev->name ), "ISAPNP" );
+	}
+	dev->parent = &rootdev->dev;
+	list_add ( &dev->siblings, &rootdev->dev.children);
+	INIT_LIST_HEAD ( &dev->children );
+
+	/* Create network device */
+	if ( ( rc = undinet_probe ( undi, dev ) ) != 0 )
+		goto err;
+
+	return 0;
+
+ err:
+	list_del ( &dev->siblings );
+	return rc;
+}
+
+/**
+ * Remove UNDI root bus
+ *
+ * @v rootdev		UNDI bus root device
+ */
+static void undibus_remove ( struct root_device *rootdev __unused ) {
+	struct undi_device *undi = &preloaded_undi;
+	struct device *dev = &undibus_dev;
+
+	undinet_remove ( undi );
+	list_del ( &dev->siblings );
+}
+
+/** UNDI bus root device driver */
+static struct root_driver undi_root_driver = {
+	.probe = undibus_probe,
+	.remove = undibus_remove,
+};
+
+/** UNDI bus root device */
+struct root_device undi_root_device __root_device = {
+	.dev = { .name = "UNDI" },
+	.driver = &undi_root_driver,
+};
+
+/**
+ * Prepare for exit
+ *
+ * @v booting		System is shutting down for OS boot
+ */
+static void undionly_shutdown ( int booting ) {
+	/* If we are shutting down to boot an OS, clear the "keep PXE
+	 * stack" flag.
+	 */
+	if ( booting )
+		preloaded_undi.flags &= ~UNDI_FL_KEEP_ALL;
+}
+
+struct startup_fn startup_undionly __startup_fn ( STARTUP_LATE ) = {
+	.name = "undionly",
+	.shutdown = undionly_shutdown,
+};

src/arch/x86/drivers/net/undipreload.c

@@ -0,0 +1,42 @@
+/*
+ * Copyright (C) 2007 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <realmode.h>
+#include <undipreload.h>
+
+/** @file
+ *
+ * Preloaded UNDI stack
+ *
+ */
+
+/**
+ * Preloaded UNDI device
+ *
+ * This is the UNDI device that was present when Etherboot started
+ * execution (i.e. when loading a .kpxe image).  The first driver to
+ * claim this device must zero out this data structure.
+ */
+struct undi_device __data16 ( preloaded_undi );

src/arch/x86/drivers/net/undirom.c

@@ -0,0 +1,235 @@
+/*
+ * Copyright (C) 2007 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER );
+
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+#include <pxe.h>
+#include <realmode.h>
+#include <undirom.h>
+
+/** @file
+ *
+ * UNDI expansion ROMs
+ *
+ */
+
+/** List of all UNDI ROMs */
+static LIST_HEAD ( undiroms );
+
+/**
+ * Parse PXE ROM ID structure
+ *
+ * @v undirom		UNDI ROM
+ * @v pxeromid		Offset within ROM to PXE ROM ID structure
+ * @ret rc		Return status code
+ */
+static int undirom_parse_pxeromid ( struct undi_rom *undirom,
+				   unsigned int pxeromid ) {
+	struct undi_rom_id undi_rom_id;
+	unsigned int undiloader;
+
+	DBGC ( undirom, "UNDIROM %p has PXE ROM ID at %04x:%04x\n", undirom,
+	       undirom->rom_segment, pxeromid );
+
+	/* Read PXE ROM ID structure and verify */
+	copy_from_real ( &undi_rom_id, undirom->rom_segment, pxeromid,
+			 sizeof ( undi_rom_id ) );
+	if ( undi_rom_id.Signature != UNDI_ROM_ID_SIGNATURE ) {
+		DBGC ( undirom, "UNDIROM %p has bad PXE ROM ID signature "
+		       "%08x\n", undirom, undi_rom_id.Signature );
+		return -EINVAL;
+	}
+
+	/* Check for UNDI loader */
+	undiloader = undi_rom_id.UNDILoader;
+	if ( ! undiloader ) {
+		DBGC ( undirom, "UNDIROM %p has no UNDI loader\n", undirom );
+		return -EINVAL;
+	}
+
+	/* Fill in UNDI ROM loader fields */
+	undirom->loader_entry.segment = undirom->rom_segment;
+	undirom->loader_entry.offset = undiloader;
+	undirom->code_size = undi_rom_id.CodeSize;
+	undirom->data_size = undi_rom_id.DataSize;
+
+	DBGC ( undirom, "UNDIROM %p has UNDI loader at %04x:%04x "
+	       "(code %04zx data %04zx)\n", undirom,
+	       undirom->loader_entry.segment, undirom->loader_entry.offset,
+	       undirom->code_size, undirom->data_size );
+	return 0;
+}
+
+/**
+ * Parse PCI expansion header
+ *
+ * @v undirom		UNDI ROM
+ * @v pcirheader	Offset within ROM to PCI expansion header
+ */
+static int undirom_parse_pcirheader ( struct undi_rom *undirom,
+				     unsigned int pcirheader ) {
+	struct pcir_header pcir_header;
+
+	DBGC ( undirom, "UNDIROM %p has PCI expansion header at %04x:%04x\n",
+	       undirom, undirom->rom_segment, pcirheader );
+
+	/* Read PCI expansion header and verify */
+	copy_from_real ( &pcir_header, undirom->rom_segment, pcirheader,
+			 sizeof ( pcir_header ) );
+	if ( pcir_header.signature != PCIR_SIGNATURE ) {
+		DBGC ( undirom, "UNDIROM %p has bad PCI expansion header "
+		       "signature %08x\n", undirom, pcir_header.signature );
+		return -EINVAL;
+	}
+
+	/* Fill in UNDI ROM PCI device fields */
+	undirom->bus_type = PCI_NIC;
+	undirom->bus_id.pci.vendor_id = pcir_header.vendor_id;
+	undirom->bus_id.pci.device_id = pcir_header.device_id;
+
+	DBGC ( undirom, "UNDIROM %p is for PCI devices %04x:%04x\n", undirom,
+	       undirom->bus_id.pci.vendor_id, undirom->bus_id.pci.device_id );
+	return 0;
+	
+}
+
+/**
+ * Probe UNDI ROM
+ *
+ * @v rom_segment	ROM segment address
+ * @ret rc		Return status code
+ */
+static int undirom_probe ( unsigned int rom_segment ) {
+	struct undi_rom *undirom = NULL;
+	struct undi_rom_header romheader;
+	size_t rom_len;
+	unsigned int pxeromid;
+	unsigned int pcirheader;
+	int rc;
+
+	/* Read expansion ROM header and verify */
+	copy_from_real ( &romheader, rom_segment, 0, sizeof ( romheader ) );
+	if ( romheader.Signature != ROM_SIGNATURE ) {
+		rc = -EINVAL;
+		goto err;
+	}
+	rom_len = ( romheader.ROMLength * 512 );
+
+	/* Allocate memory for UNDI ROM */
+	undirom = zalloc ( sizeof ( *undirom ) );
+	if ( ! undirom ) {
+		DBG ( "Could not allocate UNDI ROM structure\n" );
+		rc = -ENOMEM;
+		goto err;
+	}
+	DBGC ( undirom, "UNDIROM %p trying expansion ROM at %04x:0000 "
+	       "(%zdkB)\n", undirom, rom_segment, ( rom_len / 1024 ) );
+	undirom->rom_segment = rom_segment;
+
+	/* Check for and parse PXE ROM ID */
+	pxeromid = romheader.PXEROMID;
+	if ( ! pxeromid ) {
+		DBGC ( undirom, "UNDIROM %p has no PXE ROM ID\n", undirom );
+		rc = -EINVAL;
+		goto err;
+	}
+	if ( pxeromid > rom_len ) {
+		DBGC ( undirom, "UNDIROM %p PXE ROM ID outside ROM\n",
+		       undirom );
+		rc = -EINVAL;
+		goto err;
+	}
+	if ( ( rc = undirom_parse_pxeromid ( undirom, pxeromid ) ) != 0 )
+		goto err;
+
+	/* Parse PCIR header, if present */
+	pcirheader = romheader.PCIRHeader;
+	if ( pcirheader )
+		undirom_parse_pcirheader ( undirom, pcirheader );
+
+	/* Add to UNDI ROM list and return */
+	DBGC ( undirom, "UNDIROM %p registered\n", undirom );
+	list_add_tail ( &undirom->list, &undiroms );
+	return 0;
+
+ err:
+	free ( undirom );
+	return rc;
+}
+
+/**
+ * Create UNDI ROMs for all possible expansion ROMs
+ *
+ * @ret 
+ */
+static void undirom_probe_all_roms ( void ) {
+	static int probed = 0;
+	unsigned int rom_segment;
+
+	/* Perform probe only once */
+	if ( probed )
+		return;
+
+	DBG ( "Scanning for PXE expansion ROMs\n" );
+
+	/* Scan through expansion ROM region at 512 byte intervals */
+	for ( rom_segment = 0xc000 ; rom_segment < 0x10000 ;
+	      rom_segment += 0x20 ) {
+		undirom_probe ( rom_segment );
+	}
+
+	probed = 1;
+}
+
+/**
+ * Find UNDI ROM for PCI device
+ *
+ * @v vendor_id		PCI vendor ID
+ * @v device_id		PCI device ID
+ * @v rombase		ROM base address, or 0 for any
+ * @ret undirom		UNDI ROM, or NULL
+ */
+struct undi_rom * undirom_find_pci ( unsigned int vendor_id,
+				     unsigned int device_id,
+				     unsigned int rombase ) {
+	struct undi_rom *undirom;
+
+	undirom_probe_all_roms();
+
+	list_for_each_entry ( undirom, &undiroms, list ) {
+		if ( undirom->bus_type != PCI_NIC )
+			continue;
+		if ( undirom->bus_id.pci.vendor_id != vendor_id )
+			continue;
+		if ( undirom->bus_id.pci.device_id != device_id )
+			continue;
+		if ( rombase && ( ( undirom->rom_segment << 4 ) != rombase ) )
+			continue;
+		DBGC ( undirom, "UNDIROM %p matched PCI %04x:%04x (%08x)\n",
+		       undirom, vendor_id, device_id, rombase );
+		return undirom;
+	}
+
+	DBG ( "No UNDI ROM matched PCI %04x:%04x (%08x)\n",
+	      vendor_id, device_id, rombase );
+	return NULL;
+}

src/arch/x86/drivers/xen/hvm.c

@@ -0,0 +1,505 @@
+/*
+ * Copyright (C) 2014 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+#include <stdio.h>
+#include <errno.h>
+#include <ipxe/malloc.h>
+#include <ipxe/pci.h>
+#include <ipxe/cpuid.h>
+#include <ipxe/msr.h>
+#include <ipxe/xen.h>
+#include <ipxe/xenver.h>
+#include <ipxe/xenmem.h>
+#include <ipxe/xenstore.h>
+#include <ipxe/xenbus.h>
+#include <ipxe/xengrant.h>
+#include "hvm.h"
+
+/** @file
+ *
+ * Xen HVM driver
+ *
+ */
+
+/**
+ * Get CPUID base
+ *
+ * @v hvm		HVM device
+ * @ret rc		Return status code
+ */
+static int hvm_cpuid_base ( struct hvm_device *hvm ) {
+	struct {
+		uint32_t ebx;
+		uint32_t ecx;
+		uint32_t edx;
+	} __attribute__ (( packed )) signature;
+	uint32_t base;
+	uint32_t version;
+	uint32_t discard_eax;
+	uint32_t discard_ebx;
+	uint32_t discard_ecx;
+	uint32_t discard_edx;
+
+	/* Scan for magic signature */
+	for ( base = HVM_CPUID_MIN ; base <= HVM_CPUID_MAX ;
+	      base += HVM_CPUID_STEP ) {
+		cpuid ( base, 0, &discard_eax, &signature.ebx, &signature.ecx,
+			&signature.edx );
+		if ( memcmp ( &signature, HVM_CPUID_MAGIC,
+			      sizeof ( signature ) ) == 0 ) {
+			hvm->cpuid_base = base;
+			cpuid ( ( base + HVM_CPUID_VERSION ), 0, &version,
+				&discard_ebx, &discard_ecx, &discard_edx );
+			DBGC2 ( hvm, "HVM using CPUID base %#08x (v%d.%d)\n",
+				base, ( version >> 16 ), ( version & 0xffff ) );
+			return 0;
+		}
+	}
+
+	DBGC ( hvm, "HVM could not find hypervisor\n" );
+	return -ENODEV;
+}
+
+/**
+ * Map hypercall page(s)
+ *
+ * @v hvm		HVM device
+ * @ret rc		Return status code
+ */
+static int hvm_map_hypercall ( struct hvm_device *hvm ) {
+	uint32_t pages;
+	uint32_t msr;
+	uint32_t discard_ecx;
+	uint32_t discard_edx;
+	physaddr_t hypercall_phys;
+	uint32_t version;
+	static xen_extraversion_t extraversion;
+	int xenrc;
+	int rc;
+
+	/* Get number of hypercall pages and MSR to use */
+	cpuid ( ( hvm->cpuid_base + HVM_CPUID_PAGES ), 0, &pages, &msr,
+		&discard_ecx, &discard_edx );
+
+	/* Allocate pages */
+	hvm->hypercall_len = ( pages * PAGE_SIZE );
+	hvm->xen.hypercall = malloc_dma ( hvm->hypercall_len, PAGE_SIZE );
+	if ( ! hvm->xen.hypercall ) {
+		DBGC ( hvm, "HVM could not allocate %d hypercall page(s)\n",
+		       pages );
+		return -ENOMEM;
+	}
+	hypercall_phys = virt_to_phys ( hvm->xen.hypercall );
+	DBGC2 ( hvm, "HVM hypercall page(s) at [%#08lx,%#08lx) via MSR %#08x\n",
+		hypercall_phys, ( hypercall_phys + hvm->hypercall_len ), msr );
+
+	/* Write to MSR */
+	wrmsr ( msr, hypercall_phys );
+
+	/* Check that hypercall mechanism is working */
+	version = xenver_version ( &hvm->xen );
+	if ( ( xenrc = xenver_extraversion ( &hvm->xen, &extraversion ) ) != 0){
+		rc = -EXEN ( xenrc );
+		DBGC ( hvm, "HVM could not get extraversion: %s\n",
+		       strerror ( rc ) );
+		return rc;
+	}
+	DBGC2 ( hvm, "HVM found Xen version %d.%d%s\n",
+		( version >> 16 ), ( version & 0xffff ) , extraversion );
+
+	return 0;
+}
+
+/**
+ * Unmap hypercall page(s)
+ *
+ * @v hvm		HVM device
+ */
+static void hvm_unmap_hypercall ( struct hvm_device *hvm ) {
+
+	/* Free pages */
+	free_dma ( hvm->xen.hypercall, hvm->hypercall_len );
+}
+
+/**
+ * Allocate and map MMIO space
+ *
+ * @v hvm		HVM device
+ * @v space		Source mapping space
+ * @v len		Length (must be a multiple of PAGE_SIZE)
+ * @ret mmio		MMIO space address, or NULL on error
+ */
+static void * hvm_ioremap ( struct hvm_device *hvm, unsigned int space,
+			    size_t len ) {
+	struct xen_add_to_physmap add;
+	struct xen_remove_from_physmap remove;
+	unsigned int pages = ( len / PAGE_SIZE );
+	physaddr_t mmio_phys;
+	unsigned int i;
+	void *mmio;
+	int xenrc;
+	int rc;
+
+	/* Sanity check */
+	assert ( ( len % PAGE_SIZE ) == 0 );
+
+	/* Check for available space */
+	if ( ( hvm->mmio_offset + len ) > hvm->mmio_len ) {
+		DBGC ( hvm, "HVM could not allocate %zd bytes of MMIO space "
+		       "(%zd of %zd remaining)\n", len,
+		       ( hvm->mmio_len - hvm->mmio_offset ), hvm->mmio_len );
+		goto err_no_space;
+	}
+
+	/* Map this space */
+	mmio = pci_ioremap ( hvm->pci, ( hvm->mmio + hvm->mmio_offset ), len );
+	if ( ! mmio ) {
+		DBGC ( hvm, "HVM could not map MMIO space [%08lx,%08lx)\n",
+		       ( hvm->mmio + hvm->mmio_offset ),
+		       ( hvm->mmio + hvm->mmio_offset + len ) );
+		goto err_ioremap;
+	}
+	mmio_phys = virt_to_phys ( mmio );
+
+	/* Add to physical address space */
+	for ( i = 0 ; i < pages ; i++ ) {
+		add.domid = DOMID_SELF;
+		add.idx = i;
+		add.space = space;
+		add.gpfn = ( ( mmio_phys / PAGE_SIZE ) + i );
+		if ( ( xenrc = xenmem_add_to_physmap ( &hvm->xen, &add ) ) !=0){
+			rc = -EXEN ( xenrc );
+			DBGC ( hvm, "HVM could not add space %d idx %d at "
+			       "[%08lx,%08lx): %s\n", space, i,
+			       ( mmio_phys + ( i * PAGE_SIZE ) ),
+			       ( mmio_phys + ( ( i + 1 ) * PAGE_SIZE ) ),
+			       strerror ( rc ) );
+			goto err_add_to_physmap;
+		}
+	}
+
+	/* Update offset */
+	hvm->mmio_offset += len;
+
+	return mmio;
+
+	i = pages;
+ err_add_to_physmap:
+	for ( i-- ; ( signed int ) i >= 0 ; i-- ) {
+		remove.domid = DOMID_SELF;
+		add.gpfn = ( ( mmio_phys / PAGE_SIZE ) + i );
+		xenmem_remove_from_physmap ( &hvm->xen, &remove );
+	}
+	iounmap ( mmio );
+ err_ioremap:
+ err_no_space:
+	return NULL;
+}
+
+/**
+ * Unmap MMIO space
+ *
+ * @v hvm		HVM device
+ * @v mmio		MMIO space address
+ * @v len		Length (must be a multiple of PAGE_SIZE)
+ */
+static void hvm_iounmap ( struct hvm_device *hvm, void *mmio, size_t len ) {
+	struct xen_remove_from_physmap remove;
+	physaddr_t mmio_phys = virt_to_phys ( mmio );
+	unsigned int pages = ( len / PAGE_SIZE );
+	unsigned int i;
+	int xenrc;
+	int rc;
+
+	/* Unmap this space */
+	iounmap ( mmio );
+
+	/* Remove from physical address space */
+	for ( i = 0 ; i < pages ; i++ ) {
+		remove.domid = DOMID_SELF;
+		remove.gpfn = ( ( mmio_phys / PAGE_SIZE ) + i );
+		if ( ( xenrc = xenmem_remove_from_physmap ( &hvm->xen,
+							    &remove ) ) != 0 ) {
+			rc = -EXEN ( xenrc );
+			DBGC ( hvm, "HVM could not remove space [%08lx,%08lx): "
+			       "%s\n", ( mmio_phys + ( i * PAGE_SIZE ) ),
+			       ( mmio_phys + ( ( i + 1 ) * PAGE_SIZE ) ),
+			       strerror ( rc ) );
+			/* Nothing we can do about this */
+		}
+	}
+}
+
+/**
+ * Map shared info page
+ *
+ * @v hvm		HVM device
+ * @ret rc		Return status code
+ */
+static int hvm_map_shared_info ( struct hvm_device *hvm ) {
+	physaddr_t shared_info_phys;
+	int rc;
+
+	/* Map shared info page */
+	hvm->xen.shared = hvm_ioremap ( hvm, XENMAPSPACE_shared_info,
+					PAGE_SIZE );
+	if ( ! hvm->xen.shared ) {
+		rc = -ENOMEM;
+		goto err_alloc;
+	}
+	shared_info_phys = virt_to_phys ( hvm->xen.shared );
+	DBGC2 ( hvm, "HVM shared info page at [%#08lx,%#08lx)\n",
+		shared_info_phys, ( shared_info_phys + PAGE_SIZE ) );
+
+	/* Sanity check */
+	DBGC2 ( hvm, "HVM wallclock time is %d\n",
+		readl ( &hvm->xen.shared->wc_sec ) );
+
+	return 0;
+
+	hvm_iounmap ( hvm, hvm->xen.shared, PAGE_SIZE );
+ err_alloc:
+	return rc;
+}
+
+/**
+ * Unmap shared info page
+ *
+ * @v hvm		HVM device
+ */
+static void hvm_unmap_shared_info ( struct hvm_device *hvm ) {
+
+	/* Unmap shared info page */
+	hvm_iounmap ( hvm, hvm->xen.shared, PAGE_SIZE );
+}
+
+/**
+ * Map grant table
+ *
+ * @v hvm		HVM device
+ * @ret rc		Return status code
+ */
+static int hvm_map_grant ( struct hvm_device *hvm ) {
+	physaddr_t grant_phys;
+	int rc;
+
+	/* Initialise grant table */
+	if ( ( rc = xengrant_init ( &hvm->xen ) ) != 0 ) {
+		DBGC ( hvm, "HVM could not initialise grant table: %s\n",
+		       strerror ( rc ) );
+		return rc;
+	}
+
+	/* Map grant table */
+	hvm->xen.grant.table = hvm_ioremap ( hvm, XENMAPSPACE_grant_table,
+					     hvm->xen.grant.len );
+	if ( ! hvm->xen.grant.table )
+		return -ENODEV;
+
+	grant_phys = virt_to_phys ( hvm->xen.grant.table );
+	DBGC2 ( hvm, "HVM mapped grant table at [%08lx,%08lx)\n",
+		grant_phys, ( grant_phys + hvm->xen.grant.len ) );
+	return 0;
+}
+
+/**
+ * Unmap grant table
+ *
+ * @v hvm		HVM device
+ */
+static void hvm_unmap_grant ( struct hvm_device *hvm ) {
+
+	/* Unmap grant table */
+	hvm_iounmap ( hvm, hvm->xen.grant.table, hvm->xen.grant.len );
+}
+
+/**
+ * Map XenStore
+ *
+ * @v hvm		HVM device
+ * @ret rc		Return status code
+ */
+static int hvm_map_xenstore ( struct hvm_device *hvm ) {
+	uint64_t xenstore_evtchn;
+	uint64_t xenstore_pfn;
+	physaddr_t xenstore_phys;
+	char *name;
+	int xenrc;
+	int rc;
+
+	/* Get XenStore event channel */
+	if ( ( xenrc = xen_hvm_get_param ( &hvm->xen, HVM_PARAM_STORE_EVTCHN,
+					   &xenstore_evtchn ) ) != 0 ) {
+		rc = -EXEN ( xenrc );
+		DBGC ( hvm, "HVM could not get XenStore event channel: %s\n",
+		       strerror ( rc ) );
+		return rc;
+	}
+	hvm->xen.store.port = xenstore_evtchn;
+
+	/* Get XenStore PFN */
+	if ( ( xenrc = xen_hvm_get_param ( &hvm->xen, HVM_PARAM_STORE_PFN,
+					   &xenstore_pfn ) ) != 0 ) {
+		rc = -EXEN ( xenrc );
+		DBGC ( hvm, "HVM could not get XenStore PFN: %s\n",
+		       strerror ( rc ) );
+		return rc;
+	}
+	xenstore_phys = ( xenstore_pfn * PAGE_SIZE );
+
+	/* Map XenStore */
+	hvm->xen.store.intf = pci_ioremap ( hvm->pci, xenstore_phys,
+					    PAGE_SIZE );
+	if ( ! hvm->xen.store.intf ) {
+		DBGC ( hvm, "HVM could not map XenStore at [%08lx,%08lx)\n",
+		       xenstore_phys, ( xenstore_phys + PAGE_SIZE ) );
+		return -ENODEV;
+	}
+	DBGC2 ( hvm, "HVM mapped XenStore at [%08lx,%08lx) with event port "
+		"%d\n", xenstore_phys, ( xenstore_phys + PAGE_SIZE ),
+		hvm->xen.store.port );
+
+	/* Check that XenStore is working */
+	if ( ( rc = xenstore_read ( &hvm->xen, &name, "name", NULL ) ) != 0 ) {
+		DBGC ( hvm, "HVM could not read domain name: %s\n",
+		       strerror ( rc ) );
+		return rc;
+	}
+	DBGC2 ( hvm, "HVM running in domain \"%s\"\n", name );
+	free ( name );
+
+	return 0;
+}
+
+/**
+ * Unmap XenStore
+ *
+ * @v hvm		HVM device
+ */
+static void hvm_unmap_xenstore ( struct hvm_device *hvm ) {
+
+	/* Unmap XenStore */
+	iounmap ( hvm->xen.store.intf );
+}
+
+/**
+ * Probe PCI device
+ *
+ * @v pci		PCI device
+ * @ret rc		Return status code
+ */
+static int hvm_probe ( struct pci_device *pci ) {
+	struct hvm_device *hvm;
+	int rc;
+
+	/* Allocate and initialise structure */
+	hvm = zalloc ( sizeof ( *hvm ) );
+	if ( ! hvm ) {
+		rc = -ENOMEM;
+		goto err_alloc;
+	}
+	hvm->pci = pci;
+	hvm->mmio = pci_bar_start ( pci, HVM_MMIO_BAR );
+	hvm->mmio_len = pci_bar_size ( pci, HVM_MMIO_BAR );
+	DBGC2 ( hvm, "HVM has MMIO space [%08lx,%08lx)\n",
+		hvm->mmio, ( hvm->mmio + hvm->mmio_len ) );
+
+	/* Fix up PCI device */
+	adjust_pci_device ( pci );
+
+	/* Attach to hypervisor */
+	if ( ( rc = hvm_cpuid_base ( hvm ) ) != 0 )
+		goto err_cpuid_base;
+	if ( ( rc = hvm_map_hypercall ( hvm ) ) != 0 )
+		goto err_map_hypercall;
+	if ( ( rc = hvm_map_shared_info ( hvm ) ) != 0 )
+		goto err_map_shared_info;
+	if ( ( rc = hvm_map_grant ( hvm ) ) != 0 )
+		goto err_map_grant;
+	if ( ( rc = hvm_map_xenstore ( hvm ) ) != 0 )
+		goto err_map_xenstore;
+
+	/* Probe Xen devices */
+	if ( ( rc = xenbus_probe ( &hvm->xen, &pci->dev ) ) != 0 ) {
+		DBGC ( hvm, "HVM could not probe Xen bus: %s\n",
+		       strerror ( rc ) );
+		goto err_xenbus_probe;
+	}
+
+	pci_set_drvdata ( pci, hvm );
+	return 0;
+
+	xenbus_remove ( &hvm->xen, &pci->dev );
+ err_xenbus_probe:
+	hvm_unmap_xenstore ( hvm );
+ err_map_xenstore:
+	hvm_unmap_grant ( hvm );
+ err_map_grant:
+	hvm_unmap_shared_info ( hvm );
+ err_map_shared_info:
+	hvm_unmap_hypercall ( hvm );
+ err_map_hypercall:
+ err_cpuid_base:
+	free ( hvm );
+ err_alloc:
+	return rc;
+}
+
+/**
+ * Remove PCI device
+ *
+ * @v pci		PCI device
+ */
+static void hvm_remove ( struct pci_device *pci ) {
+	struct hvm_device *hvm = pci_get_drvdata ( pci );
+
+	xenbus_remove ( &hvm->xen, &pci->dev );
+	hvm_unmap_xenstore ( hvm );
+	hvm_unmap_grant ( hvm );
+	hvm_unmap_shared_info ( hvm );
+	hvm_unmap_hypercall ( hvm );
+	free ( hvm );
+}
+
+/** PCI device IDs */
+static struct pci_device_id hvm_ids[] = {
+	PCI_ROM ( 0x5853, 0x0001, "hvm", "hvm", 0 ),
+	PCI_ROM ( 0x5853, 0x0002, "hvm2", "hvm2", 0 ),
+};
+
+/** PCI driver */
+struct pci_driver hvm_driver __pci_driver = {
+	.ids = hvm_ids,
+	.id_count = ( sizeof ( hvm_ids ) / sizeof ( hvm_ids[0] ) ),
+	.probe = hvm_probe,
+	.remove = hvm_remove,
+};
+
+/* Drag in objects via hvm_driver */
+REQUIRING_SYMBOL ( hvm_driver );
+
+/* Drag in netfront driver */
+REQUIRE_OBJECT ( netfront );

src/arch/x86/drivers/xen/hvm.h

@@ -0,0 +1,77 @@
+#ifndef _HVM_H
+#define _HVM_H
+
+/** @file
+ *
+ * Xen HVM driver
+ *
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+#include <stdint.h>
+#include <ipxe/xen.h>
+#include <xen/hvm/hvm_op.h>
+#include <xen/hvm/params.h>
+
+/** Minimum CPUID base */
+#define HVM_CPUID_MIN 0x40000000UL
+
+/** Maximum CPUID base */
+#define HVM_CPUID_MAX 0x4000ff00UL
+
+/** Increment between CPUID bases */
+#define HVM_CPUID_STEP 0x00000100UL
+
+/** Magic signature */
+#define HVM_CPUID_MAGIC "XenVMMXenVMM"
+
+/** Get Xen version */
+#define HVM_CPUID_VERSION 1
+
+/** Get number of hypercall pages */
+#define HVM_CPUID_PAGES 2
+
+/** PCI MMIO BAR */
+#define HVM_MMIO_BAR PCI_BASE_ADDRESS_1
+
+/** A Xen HVM device */
+struct hvm_device {
+	/** Xen hypervisor */
+	struct xen_hypervisor xen;
+	/** PCI device */
+	struct pci_device *pci;
+	/** CPUID base */
+	uint32_t cpuid_base;
+	/** Length of hypercall table */
+	size_t hypercall_len;
+	/** MMIO base address */
+	unsigned long mmio;
+	/** Current offset within MMIO address space */
+	size_t mmio_offset;
+	/** Length of MMIO address space */
+	size_t mmio_len;
+};
+
+/**
+ * Get HVM parameter value
+ *
+ * @v xen		Xen hypervisor
+ * @v index		Parameter index
+ * @v value		Value to fill in
+ * @ret xenrc		Xen status code
+ */
+static inline int xen_hvm_get_param ( struct xen_hypervisor *xen,
+				      unsigned int index, uint64_t *value ) {
+	struct xen_hvm_param param;
+	int xenrc;
+
+	param.domid = DOMID_SELF;
+	param.index = index;
+	xenrc = xen_hypercall_2 ( xen, __HYPERVISOR_hvm_op, HVMOP_get_param,
+				  virt_to_phys ( &param ) );
+	*value = param.value;
+	return xenrc;
+}
+
+#endif /* _HVM_H */

src/arch/x86/hci/commands/pxe_cmd.c

@@ -0,0 +1,117 @@
+/*
+ * Copyright (C) 2010 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+#include <ipxe/netdevice.h>
+#include <ipxe/command.h>
+#include <ipxe/parseopt.h>
+#include <hci/ifmgmt_cmd.h>
+#include <pxe_call.h>
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/** @file
+ *
+ * PXE commands
+ *
+ */
+
+/** "startpxe" options */
+struct startpxe_options {};
+
+/** "startpxe" option list */
+static struct option_descriptor startpxe_opts[] = {};
+
+/**
+ * "startpxe" payload
+ *
+ * @v netdev		Network device
+ * @v opts		Command options
+ * @ret rc		Return status code
+ */
+static int startpxe_payload ( struct net_device *netdev,
+			      struct startpxe_options *opts __unused ) {
+
+	if ( netdev_is_open ( netdev ) )
+		pxe_activate ( netdev );
+
+	return 0;
+}
+
+/** "startpxe" command descriptor */
+static struct ifcommon_command_descriptor startpxe_cmd =
+	IFCOMMON_COMMAND_DESC ( struct startpxe_options, startpxe_opts,
+				0, MAX_ARGUMENTS, "[<interface>]",
+				startpxe_payload, 0 );
+
+/**
+ * The "startpxe" command
+ *
+ * @v argc		Argument count
+ * @v argv		Argument list
+ * @ret rc		Return status code
+ */
+static int startpxe_exec ( int argc, char **argv ) {
+	return ifcommon_exec ( argc, argv, &startpxe_cmd );
+}
+
+/** "stoppxe" options */
+struct stoppxe_options {};
+
+/** "stoppxe" option list */
+static struct option_descriptor stoppxe_opts[] = {};
+
+/** "stoppxe" command descriptor */
+static struct command_descriptor stoppxe_cmd =
+	COMMAND_DESC ( struct stoppxe_options, stoppxe_opts, 0, 0, NULL );
+
+/**
+ * The "stoppxe" command
+ *
+ * @v argc		Argument count
+ * @v argv		Argument list
+ * @ret rc		Return status code
+ */
+static int stoppxe_exec ( int argc __unused, char **argv __unused ) {
+	struct stoppxe_options opts;
+	int rc;
+
+	/* Parse options */
+	if ( ( rc = parse_options ( argc, argv, &stoppxe_cmd, &opts ) ) != 0 )
+		return rc;
+
+	pxe_deactivate();
+
+	return 0;
+}
+
+/** PXE commands */
+struct command pxe_commands[] __command = {
+	{
+		.name = "startpxe",
+		.exec = startpxe_exec,
+	},
+	{
+		.name = "stoppxe",
+		.exec = stoppxe_exec,
+	},
+};

src/arch/x86/image/bootsector.c

@@ -0,0 +1,141 @@
+/*
+ * Copyright (C) 2007 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of the
+ * License, or any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
+ * 02110-1301, USA.
+ *
+ * You can also choose to distribute this program under the terms of
+ * the Unmodified Binary Distribution Licence (as given in the file
+ * COPYING.UBDL), provided that you have satisfied its requirements.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
+
+/**
+ * @file
+ *
+ * x86 bootsector image format
+ *
+ */
+
+#include <errno.h>
+#include <realmode.h>
+#include <biosint.h>
+#include <bootsector.h>
+#include <ipxe/console.h>
+
+/** Vector for storing original INT 18 handler
+ *
+ * We do not chain to this vector, so there is no need to place it in
+ * .text16.
+ */
+static struct segoff int18_vector;
+
+/** Vector for storing original INT 19 handler
+ *
+ * We do not chain to this vector, so there is no need to place it in
+ * .text16.
+ */
+static struct segoff int19_vector;
+
+/** Restart point for INT 18 or 19 */
+extern void bootsector_exec_fail ( void );
+
+/**
+ * Jump to preloaded bootsector
+ *
+ * @v segment		Real-mode segment
+ * @v offset		Real-mode offset
+ * @v drive		Drive number to pass to boot sector
+ * @ret rc		Return status code
+ */
+int call_bootsector ( unsigned int segment, unsigned int offset,
+		      unsigned int drive ) {
+	int discard_b, discard_D, discard_d;
+
+	/* Reset console, since boot sector will probably use it */
+	console_reset();
+
+	DBG ( "Booting from boot sector at %04x:%04x\n", segment, offset );
+
+	/* Hook INTs 18 and 19 to capture failure paths */
+	hook_bios_interrupt ( 0x18, ( intptr_t ) bootsector_exec_fail,
+			      &int18_vector );
+	hook_bios_interrupt ( 0x19, ( intptr_t ) bootsector_exec_fail,
+			      &int19_vector );
+
+	/* Boot the loaded sector
+	 *
+	 * We assume that the boot sector may completely destroy our
+	 * real-mode stack, so we preserve everything we need in
+	 * static storage.
+	 */
+	__asm__ __volatile__ ( REAL_CODE ( /* Save return address off-stack */
+					   "popw %%cs:saved_retaddr\n\t"
+					   /* Save stack pointer */
+					   "movw %%ss, %%ax\n\t"
+					   "movw %%ax, %%cs:saved_ss\n\t"
+					   "movw %%sp, %%cs:saved_sp\n\t"
+					   /* Save frame pointer (gcc bug) */
+					   "movl %%ebp, %%cs:saved_ebp\n\t"
+					   /* Prepare jump to boot sector */
+					   "pushw %%bx\n\t"
+					   "pushw %%di\n\t"
+					   /* Clear all registers */
+					   "xorl %%eax, %%eax\n\t"
+					   "xorl %%ebx, %%ebx\n\t"
+					   "xorl %%ecx, %%ecx\n\t"
+					   /* %edx contains drive number */
+					   "xorl %%esi, %%esi\n\t"
+					   "xorl %%edi, %%edi\n\t"
+					   "xorl %%ebp, %%ebp\n\t"
+					   "movw %%ax, %%ds\n\t"
+					   "movw %%ax, %%es\n\t"
+					   "movw %%ax, %%fs\n\t"
+					   "movw %%ax, %%gs\n\t"
+					   /* Jump to boot sector */
+					   "sti\n\t"
+					   "lret\n\t"
+					   /* Preserved variables */
+					   "\nsaved_ebp: .long 0\n\t"
+					   "\nsaved_ss: .word 0\n\t"
+					   "\nsaved_sp: .word 0\n\t"
+					   "\nsaved_retaddr: .word 0\n\t"
+					   /* Boot failure return point */
+					   "\nbootsector_exec_fail:\n\t"
+					   /* Restore frame pointer (gcc bug) */
+					   "movl %%cs:saved_ebp, %%ebp\n\t"
+					   /* Restore stack pointer */
+					   "movw %%cs:saved_ss, %%ax\n\t"
+					   "movw %%ax, %%ss\n\t"
+					   "movw %%cs:saved_sp, %%sp\n\t"
+					   /* Return via saved address */
+					   "jmp *%%cs:saved_retaddr\n\t" )
+			       : "=b" ( discard_b ), "=D" ( discard_D ),
+			         "=d" ( discard_d )
+			       : "b" ( segment ), "D" ( offset ),
+			         "d" ( drive )
+			       : "eax", "ecx", "esi" );
+
+	DBG ( "Booted disk returned via INT 18 or 19\n" );
+
+	/* Unhook INTs 18 and 19 */
+	unhook_bios_interrupt ( 0x18, ( intptr_t ) bootsector_exec_fail,
+				&int18_vector );
+	unhook_bios_interrupt ( 0x19, ( intptr_t ) bootsector_exec_fail,
+				&int19_vector );
+	
+	return -ECANCELED;
+}