vmmon: avoid using alloc_vm_area() for crosspage allocation

Mainline commit 301fa9f2ddf7 ("mm: remove alloc_vm_area") in v5.10-rc1 removed alloc_vm_area() function used to allow allocating crosspage as executable after vmap() started silently setting NX flag in kernel 5.8. Similar to what upstream did in Xen code, use apply_to_page_range() to fix the crosspage flags, except we don't need to use get_vm_area() which is not exported. Thanks to virtualbox for the hint that current->mm can be used instead of init_mm (which is not exported either).
vmmon: avoid kernel writes to eventfd
2020-12-09 11:20:45 +01:00 · 2020-12-09 11:20:45 +01:00 · 2020-12-09 11:20:45 +01:00 · 2020-12-09 11:20:45 +01:00 · 2020-12-09 11:20:45 +01:00 · 2020-12-09 11:20:45 +01:00
45 changed files with 1124 additions and 2702 deletions
@@ -3,6 +3,8 @@
 *.ko
 *.ko.cmd
 *.mod.c
+*.tar
 .tmp_versions
+.cache.mk
 Module.symvers
 modules.order
@@ -0,0 +1,197 @@
+This document explains how to use the repository to retrieve the module
+source, build modules and install them. There are two basic methods: we can
+either build the modules from source and install them ourselves or replace
+VMware provided source tarballs with patched ones and let it use its own
+vmware-modconfig tool.
+
+
+0. Quick guide for impatient
+----------------------------
+
+First method (build and install):
+
+  wget https://github.com/mkubecek/vmware-host-modules/archive/workstation-14.1.1.tar.gz
+  tar -xzf workstation-14.1.1.tar.gz
+  cd vmware-host-modules-workstation-14.1.1
+  make
+  make install
+
+Last command must be executed with root privileges (first four shouldn't).
+Based on your VMware product, replace "14.1.1" with your installed version
+and/or "workstation" with "player".
+
+Second method (replace original tarballs):
+
+  wget https://github.com/mkubecek/vmware-host-modules/archive/workstation-14.1.1.tar.gz
+  tar -xzf workstation-14.1.1.tar.gz
+  cd vmware-host-modules-workstation-14.1.1
+  tar -cf vmmon.tar vmmon-only
+  tar -cf vmnet.tar vmnet-only
+  cp -v vmmon.tar vmnet.tar /usr/lib/vmware/modules/source/
+  vmware-modconfig --console --install-all
+
+In this case, last two commands require root privileges.
+
+
+1a. Retrieve the source - with git
+----------------------------------
+
+Using git is preferrable as it allows an easy switch to sources for
+different product/version or update for newer kernel version. Without git,
+you can only download one particular source snapshot.
+
+First, clone the repository from GitHub either using HTTPS
+
+  git clone https://github.com/mkubecek/vmware-host-modules.git
+
+or using SSH
+
+  git clone git@github.com:mkubecek/vmware-host-modules.git
+
+Using SSH is preferrable if you already have an account on GitHub, use
+HTTPS if you don't.
+
+Branch "master" cannot be used to build modules, it contains only common
+files so that changes in them can be merged into all other branches easily.
+To get actual sources, checkout a "real" branch, e.g.
+
+  git checkout workstation-14.1.1
+
+Branch name consists of "workstation" (for Workstation) or "player" (for
+Player), a dash and version number ("14.1.1" here). Always use correct
+product and version, the modules check if internal source version matches
+installed product and refuse to load otherwise. Actually, in all recent
+versions, "workstation" and "player" sources have been identical but there
+is no guarantee it will always be the case.
+
+To pull out updated branch (if there are new commits), run "git pull", to
+switch to a new branch run "git fetch" and "git checkout ...".
+
+
+1b. Retrieve the source - without git
+-------------------------------------
+
+GitHub allows to download a particular snapshot (branch or tag) directly
+without git:
+
+  wget https://github.com/mkubecek/vmware-host-modules/archive/workstation-14.1.1.tar.gz
+
+where "workstation" and "14.1.1" have the same meaning as in section 1a.
+Any other tool can be used to retrieve the URL, of course. It's also
+possible to download directly from the web interface, switch to the branch
+you are interested in, click the "Clone or download" button and then
+"Download ZIP".
+
+Unpack the downloaded archive
+
+  tar -xzf workstation-14.1.1.tar.gz
+
+or
+
+  unzip vmware-host-modules-workstation-14.1.1.zip
+
+depending on archive type. Then change to the resulting directory.
+
+
+2a. Build and install modules - directly
+----------------------------------------
+
+To build modules against currently running kernel, just run "make" in top
+level directory. It creates two module files vmmon-only/vmmon.ko and
+vmnet-only/vmnet.ko (unstripped). To build against a different kernel, set
+variable VM_UNAME, e.g.
+
+  make VM_UNAME='4.14.15-5-default'
+
+As name suggests, this variable should be set to what "uname -r" would show
+when that kernel is booted (it's the same as name of the subdirectory under
+/lib/modules with modules of that kernel).
+
+To install built modules for currently running kernel, run "make install".
+This command needs to be run as user with permissions to write into the
+directory with kernel modules (under normal circumstances, this should be
+only root). The VM_UNAME variable can be used to install for a different
+kernel. Unless you know well what are you doing and why, the same value
+should be always used for "make" and "make install". That's why makefile
+checks if vermagic tag of the moudules match VM_UNAME (or current kernel if
+VM_UNAME isn't set).
+
+If DESTDIR is not set (or is empty), "make install" also runs depmod to
+update module dependency cache.
+
+Note that "make install" replaces existing module files but it does not
+replace modules in running kernel if they are already loaded. Reloading the
+modules can be enforced e.g. by
+
+  /etc/init.d/vmware restart
+
+All running VMs should be turned off before this.
+
+
+2b. Build and install modules - replacing the original source tarballs
+----------------------------------------------------------------------
+
+Manual build and installation of host modules may be natural for developers
+and power users but those who prefer to "just use" would appreciate if they
+could rely on standard VMware tools. To this goal, we need to replace the
+module source tarballs with patched versions.
+
+In local checked out repository, the replacement tarballs can be created
+with
+
+  make tarballs
+
+This command creates files vmmon.tar and vmnet.tar which can be used to
+replace the original tarballs provided with the VMware product. Note that
+this command creates the tarball from current HEAD (in git language), i.e.
+they do not contain any uncommited local changes (e.g. files created by
+test build).
+
+When using a downloaded tarball, simply run
+
+  tar -cf vmmon.tar vmmon-only
+  tar -cf vmnet.tar vmnet-only
+
+to create the tarballs. In this case, there is no protection against
+unwanted local changes.
+
+Whatever way you used to create the tarballs, replace the original ones
+provided by VMware
+
+  /usr/lib/vmware/modules/source/vmmon.tar
+  /usr/lib/vmware/modules/source/vmnet.tar
+
+by patched versions. It is highly recommended to backup the original
+tarballs before replacing them.
+
+Once patched tarballs are installed, you can rebuild the modules as usual:
+
+  vmware-modconfig --console --install-all
+
+or let the GUI run the command for you.
+
+Replacing the original tarballs prevents having to rebuild and install the
+modules manually with every new kernel version. However, once the VMware
+product is upgraded or reinstalled, tarballs are replaced so that if the
+new VMware version doesn't build/work with a recent kernel, you still need
+to repeat the process. The same also holds if there is a new kernel which
+requires updated version of the patches.
+
+
+3. Final notes
+--------------
+
+Always make sure that you are using a branch matching the Workstation or
+Player version you are running. VMware software may refuse to start if
+module intended for different version is loaded. And unless you have
+a reason not to, use current head of that branch; even if the module
+builds, there may still be later fixes which may be important.
+
+In some web forums, one can comments suggesting to rip out one patched file
+and use it as replacement for the corresponding file with VMware tarballs.
+Please don't do this - and if you do, don't complain if something doesn't
+work. A commit fixing some problem is meant to be used as a whole; picking
+just changes in one file can lead to rather unpredictable results. While
+picking just one commit (or some of the commits) in a branch is perfectly
+fine (you should still know what are you doing and why), picking just one
+patched file rarely makes sense.
@@ -0,0 +1,49 @@
+MODULES = vmmon vmnet
+SUBDIRS = $(MODULES:%=%-only)
+TARBALLS = $(MODULES:%=%.tar)
+MODFILES = $(foreach mod,$(MODULES),$(mod)-only/$(mod).ko)
+VM_UNAME = $(shell uname -r)
+MODDIR = /lib/modules/$(VM_UNAME)/misc
+
+MODINFO = /sbin/modinfo
+DEPMOD = /sbin/depmod
+
+%.tar: FORCE gitcleancheck
+	git archive -o $@ --format=tar HEAD $(@:.tar=-only)
+
+.PHONY: FORCE subdirs $(SUBDIRS) clean tarballs
+
+subdirs: retiredcheck $(SUBDIRS)
+
+FORCE:
+
+$(SUBDIRS):
+	$(MAKE) -C $@ $(MAKECMDGOALS)
+
+gitcheck:
+	@git status >/dev/null 2>&1 \
+	     || ( echo "This only works in a git repository."; exit 1 )
+
+gitcleancheck: gitcheck
+	@git diff --exit-code HEAD >/dev/null 2>&1 \
+	     || echo "Warning: tarballs will reflect current HEAD (no uncommited changes)"
+
+retiredcheck:
+	@test -f RETIRED && cat RETIRED || true
+
+install: retiredcheck $(MODFILES)
+	@for f in $(MODFILES); do \
+	    mver=$$($(MODINFO) -F vermagic $$f);\
+	    mver=$${mver%% *};\
+	    test "$${mver}" = "$(VM_UNAME)" \
+	        || ( echo "Version mismatch: module $$f $${mver}, kernel $(VM_UNAME)" ; exit 1 );\
+	done
+	install -D -t $(DESTDIR)$(MODDIR) $(MODFILES)
+	strip --strip-debug $(MODULES:%=$(DESTDIR)$(MODDIR)/%.ko)
+	if test -z "$(DESTDIR)"; then $(DEPMOD) -a $(VM_UNAME); fi
+
+clean: $(SUBDIRS)
+	rm -f *.o
+
+tarballs: $(TARBALLS)
+
@@ -12,12 +12,17 @@ a particular version of Player or Workstation.

 From these tags, branches "player-${version}" and "workstation-${version}"
 are forked. These branches track changes needed to build these modules
-against recent kernel versions. Tag "p${ver}-k${ver}" marks the set of
-changes needed to build modules from Player ${ver} against kernel ${ver},
+against recent kernel versions. Tag "p${ver}-k${kver}" marks the set of
+changes needed to build modules from Player ${ver} against kernel ${kver},
 e.g. p12.5.5-k4.11 is for Player 12.5.5 and kernel 4.11; similar naming
 scheme "w${ver}-k${kver}" is used for Workstation modules. In general,
 later points in the branch should also work with older kernels (e.g.
-p12.5.5-k4.11 with kernel 4.10) but it's not completely reliable.
+p12.5.5-k4.11 with kernel 4.10) but it's not guaranteed; it may happen that
+a fix for new kernel version may be too complicated to make also work with
+older versions.
+
+At the moment, changes are tested to build against all (vanilla) kernel
+releases starting with 4.9.

 This repository is provided "as is" with no guarantees. Use the contents on
 your own risk.
@@ -117,7 +117,7 @@ prebuild:: ;
 postbuild:: ;

 $(DRIVER_KO): prebuild
-	$(MAKE) -C $(BUILD_DIR) SUBDIRS=$$PWD SRCROOT=$$PWD/$(SRCROOT) \
+	$(MAKE) -C $(BUILD_DIR) M=$$PWD SRCROOT=$$PWD/$(SRCROOT) \
 	  MODULEBUILDDIR=$(MODULEBUILDDIR) modules
 	$(MAKE) -C $$PWD SRCROOT=$$PWD/$(SRCROOT) \
 	  MODULEBUILDDIR=$(MODULEBUILDDIR) postbuild
@@ -33,7 +33,7 @@ $(DRIVER)-y := $(subst $(SRCROOT)/, , $(patsubst %.c, %.o, \
 		$(wildcard $(SRCROOT)/linux/*.c $(SRCROOT)/common/*.c $(SRCROOT)/vmcore/*.c)))

 clean:
-	rm -rf $(wildcard $(DRIVER).mod.c $(DRIVER).ko .tmp_versions \
+	rm -rf $(wildcard $(DRIVER).mod.c $(DRIVER).ko .tmp_versions .cache.mk \
 		Module.symvers Modules.symvers Module.markers modules.order \
 		$(foreach dir,linux/ common/ vmcore/ \
 		./,$(addprefix $(dir),.*.cmd .*.o.flags *.o)))
@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998,2018 VMware, Inc. All rights reserved.
 *
 * 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
@@ -35,6 +35,7 @@
 uint32 cpuidFeatures;
 static CpuidVendor vendor = CPUID_NUM_VENDORS;
 static uint32 version;
+static Bool hasSpecCtrl;


 /*
@@ -97,6 +98,9 @@ CPUID_Init(void)
      Warning("VMMON CPUID: Unrecognized CPU\n");
      vendor = CPUID_VENDOR_UNKNOWN;
   }
+
+   __GET_CPUID2(7, 0, &regs);
+   hasSpecCtrl = vendor == CPUID_VENDOR_INTEL && ((regs.edx >> 26) & 0x3) != 0;
 }


@@ -207,3 +211,26 @@ CPUID_AddressSizeSupported(void)

   return result;
 }
+
+
+/*
+ *-----------------------------------------------------------------------------
+ *
+ * CPUID_HostSupportsSpecCtrl --
+ *
+ *     Determine whether the processor supports MSR_SPEC_CTRL.
+ *
+ * Results:
+ *     True iff the processor supports the required features.
+ *
+ * Side effects:
+ *     None.
+ *
+ *-----------------------------------------------------------------------------
+ */
+
+Bool
+CPUID_HostSupportsSpecCtrl(void)
+{
+   return hasSpecCtrl;
+}
@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998,2016 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998,2016,2018 VMware, Inc. All rights reserved.
 *
 * 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
@@ -39,6 +39,7 @@ uint32 CPUID_GetVersion(void);
 Bool CPUID_SyscallSupported(void);
 Bool CPUID_LongModeSupported(void);
 Bool CPUID_AddressSizeSupported(void);
+Bool CPUID_HostSupportsSpecCtrl(void);

 static INLINE uint32 
 CPUID_GetFeatures(void)
@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998-2014 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998-2017 VMware, Inc. All rights reserved.
 *
 * 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
@@ -122,14 +122,10 @@ EXTERN MPN HostIF_GetNextAnonPage(VMDriver *vm, MPN mpn);
 EXTERN int HostIF_GetLockedPageList(VMDriver *vm, VA64 uAddr,
                                    unsigned int numPages);

-EXTERN int HostIF_ReadPage(MPN mpn, VA64 addr, Bool kernelBuffer);
-EXTERN int HostIF_WritePage(MPN mpn, VA64 addr, Bool kernelBuffer);
-#ifdef _WIN32
-/* Add a HostIF_ReadMachinePage() if/when needed */
+EXTERN int HostIF_ReadPage(VMDriver *vm, MPN mpn, VA64 addr, Bool kernelBuffer);
+EXTERN int HostIF_WritePage(VMDriver *vm, MPN mpn, VA64 addr,
+                            Bool kernelBuffer);
 EXTERN int HostIF_WriteMachinePage(MPN mpn, VA64 addr);
-#else
-#define HostIF_WriteMachinePage(_a, _b) HostIF_WritePage((_a), (_b), TRUE)
-#endif
 #if defined __APPLE__
 // There is no need for a fast clock lock on Mac OS.
 #define HostIF_FastClockLock(_callerID) do {} while (0)
@@ -145,4 +141,8 @@ EXTERN void HostIF_FreeMachinePage(MPN mpn);

 EXTERN int HostIF_SafeRDMSR(uint32 msr, uint64 *val);

+#if defined __APPLE__
+EXTERN void HostIF_PageUnitTest(void);
+#endif
+
 #endif // ifdef _HOSTIF_H_
@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998-2017 VMware, Inc. All rights reserved.
 *
 * 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
@@ -88,6 +88,7 @@

 #include "vmware.h"
 #include "hostif.h"
+#include "vmx86.h"

 #include "memtrack.h"

@@ -146,12 +147,11 @@ typedef struct MemTrackHT {
 typedef uint64 MemTrackHTKey;

 typedef struct MemTrack {
+   VMDriver         *vm;            /* The VM instance. */
   unsigned          numPages;      /* Number of pages tracked. */
   MemTrackDir1      dir1;          /* First level directory. */
   MemTrackHT        vpnHashTable;  /* VPN to entry hashtable. */
-#if defined(MEMTRACK_MPN_LOOKUP)
   MemTrackHT        mpnHashTable;  /* MPN to entry hashtable. */
-#endif
 } MemTrack;

 /*
@@ -304,11 +304,9 @@ MemTrackCleanup(MemTrack *mt)    // IN
      if (mt->vpnHashTable.pages[idx] != NULL) {
         HostIF_FreePage(mt->vpnHashTable.pages[idx]);
      }
-#if defined(MEMTRACK_MPN_LOOKUP)
      if (mt->mpnHashTable.pages[idx] != NULL) {
         HostIF_FreePage(mt->mpnHashTable.pages[idx]);
      }
-#endif
   }

   HostIF_FreeKernelMem(mt);
@@ -332,7 +330,7 @@ MemTrackCleanup(MemTrack *mt)    // IN
 */

 MemTrack *
-MemTrack_Init(void)
+MemTrack_Init(VMDriver *vm) // IN:
 {
   MemTrack *mt;
   unsigned idx;
@@ -349,6 +347,7 @@ MemTrack_Init(void)
      goto error;
   }
   memset(mt, 0, sizeof *mt);
+   mt->vm = vm;

   for (idx = 0; idx < MEMTRACK_HT_PAGES; idx++) {
      MemTrackHTPage *htPage = MemTrackAllocPage();
@@ -360,7 +359,6 @@ MemTrack_Init(void)
      mt->vpnHashTable.pages[idx] = htPage;
   }

-#if defined(MEMTRACK_MPN_LOOKUP)
   for (idx = 0; idx < MEMTRACK_HT_PAGES; idx++) {
      MemTrackHTPage *htPage = MemTrackAllocPage();

@@ -370,7 +368,6 @@ MemTrack_Init(void)
      }
      mt->mpnHashTable.pages[idx] = htPage;
   }
-#endif

   return mt;

@@ -409,6 +406,8 @@ MemTrack_Add(MemTrack *mt,    // IN
   MemTrackDir3 *dir3;
   MEMTRACK_IDX2DIR(idx, p1, p2, p3);

+   ASSERT(HostIF_VMLockIsHeld(mt->vm));
+
   if (p1 >= MEMTRACK_DIR1_ENTRIES ||
       p2 >= MEMTRACK_DIR2_ENTRIES ||
       p3 >= MEMTRACK_DIR3_ENTRIES) {
@@ -430,9 +429,7 @@ MemTrack_Add(MemTrack *mt,    // IN
   ent->mpn = mpn;

   MemTrackHTInsert(&mt->vpnHashTable, ent, &ent->vpnChain, ent->vpn);
-#if defined(MEMTRACK_MPN_LOOKUP)
   MemTrackHTInsert(&mt->mpnHashTable, ent, &ent->mpnChain, ent->mpn);
-#endif

   mt->numPages++;

@@ -461,6 +458,7 @@ MemTrack_LookupVPN(MemTrack *mt, // IN
                   VPN64 vpn)    // IN
 {
   MemTrackEntry *next = *MemTrackHTLookup(&mt->vpnHashTable, vpn);
+   ASSERT(HostIF_VMLockIsHeld(mt->vm));

   while (next != NULL) {
      if (next->vpn == vpn) {
@@ -473,7 +471,6 @@ MemTrack_LookupVPN(MemTrack *mt, // IN
 }


-#if defined(MEMTRACK_MPN_LOOKUP)
 /*
 *----------------------------------------------------------------------
 *
@@ -493,7 +490,9 @@ MemTrackEntry *
 MemTrack_LookupMPN(MemTrack *mt, // IN
                   MPN mpn)      // IN
 {
-   MemTrackEntry *next = *MemTrackHTLookup(&mt->mpnHashTable, mpn);
+   MemTrackEntry *next;
+   ASSERT(HostIF_VMLockIsHeld(mt->vm));
+   next = *MemTrackHTLookup(&mt->mpnHashTable, mpn);

   while (next != NULL) {
      if (next->mpn == mpn) {
@@ -504,7 +503,6 @@ MemTrack_LookupMPN(MemTrack *mt, // IN

   return NULL;
 }
-#endif


 /*
@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998-2017 VMware, Inc. All rights reserved.
 *
 * 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
@@ -31,30 +31,22 @@
 #define INCLUDE_ALLOW_VMCORE
 #include "includeCheck.h"

-#if defined(VMX86_DEBUG)
-#define MEMTRACK_MPN_LOOKUP
-#endif
-
 struct MemTrack;

 typedef struct MemTrackEntry {
   VPN64                   vpn;
   MPN                     mpn;
   struct MemTrackEntry   *vpnChain;
-#if defined(MEMTRACK_MPN_LOOKUP)
   struct MemTrackEntry   *mpnChain;
-#endif
 } MemTrackEntry;

 typedef void (MemTrackCleanupCb)(void *cData, MemTrackEntry *entry);

-extern struct MemTrack *MemTrack_Init(void);
+extern struct MemTrack *MemTrack_Init(VMDriver *vm);
 extern unsigned MemTrack_Cleanup(struct MemTrack *mt, MemTrackCleanupCb *cb,
                                 void *cbData);
 extern MemTrackEntry *MemTrack_Add(struct MemTrack *mt, VPN64 vpn, MPN mpn);
 extern MemTrackEntry *MemTrack_LookupVPN(struct MemTrack *mt, VPN64 vpn);
-#if defined(MEMTRACK_MPN_LOOKUP)
 extern MemTrackEntry *MemTrack_LookupMPN(struct MemTrack *mt, MPN mpn);
-#endif

 #endif // _MEMTRACK_H_
@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998-2016 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998-2018 VMware, Inc. All rights reserved.
 *
 * 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
@@ -39,6 +39,9 @@
 #   include <linux/string.h> /* memset() in the kernel */

 #   define EXPORT_SYMTAB
+#   if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 12, 0)
+#      define LINUX_GDT_IS_RO
+#   endif
 #else
 #   include <string.h>
 #endif
@@ -60,6 +63,13 @@
 #include "apic.h"
 #include "x86perfctr.h"

+#ifdef LINUX_GDT_IS_RO
+#   include <asm/desc.h>
+#   define default_rw_gdt get_current_gdt_rw()
+#else
+#   define default_rw_gdt NULL
+#endif
+
 #if defined(_WIN64)
 #   include "x86.h"
 #   include "vmmon-asm-x86-64.h"
@@ -69,6 +79,9 @@
 #   define USE_TEMPORARY_GDT 0
 #endif

+/* Cycle through values of SPEC_CTRL and verify they survive world switch. */
+#undef CYCLE_SPEC_CTRL
+
 #define TS_ASSERT(t) do { \
   DEBUG_ONLY(if (!(t)) TaskAssertFail(__LINE__);)  \
 } while (0)
@@ -691,11 +704,28 @@ TaskRestoreHostGDTTRLDT(Descriptor *tempGDTBase,
       */

      desc = (Descriptor *)((VA)HOST_KERNEL_LA_2_VA(hostGDT64.offset + tr));
+#ifdef LINUX_GDT_IS_RO
+      /*
+       * If GDT is read-only, we must always load TR from alternative gdt,
+       * otherwise CPU gets page fault when marking TR busy.
+       */
+      {
+         DTR64 rwGDT64;
+
+         rwGDT64.offset = (unsigned long)tempGDTBase;
+         rwGDT64.limit = hostGDT64.limit;
+         Desc_SetType((Descriptor *)((unsigned long)tempGDTBase + tr), TASK_DESC);
+         _Set_GDT((DTR *)&rwGDT64);
+         SET_TR(tr);
+         _Set_GDT((DTR *)&hostGDT64);
+      }
+#else
      if (Desc_Type(desc) == TASK_DESC_BUSY) {
         Desc_SetType(desc, TASK_DESC);
      }
      _Set_GDT((DTR *)&hostGDT64);
      SET_TR(tr);
+#endif
      SET_LDT(ldt);
   }
 }
@@ -1567,12 +1597,23 @@ TaskSwitchToMonitor(VMCrossPage *crosspage)
   {
      uint64 raxGetsWiped, rcxGetsWiped;

+#ifdef CALL_NOSPEC
+      __asm__ __volatile__(CALL_NOSPEC
+                           : "=a" (raxGetsWiped),
+                             "=c" (rcxGetsWiped)
+                           : "0" (codePtr),
+                             "1" (crosspage),
+                             THUNK_TARGET(codePtr)
+                           : "rdx", "r8", "r9", "r10", "r11", "cc", "memory");
+#else
      __asm__ __volatile__("call *%%rax"
                           : "=a" (raxGetsWiped),
                             "=c" (rcxGetsWiped)
                           : "0" (codePtr),
                             "1" (crosspage)
                           : "rdx", "r8", "r9", "r10", "r11", "cc", "memory");
+#endif
+
   }
 #elif defined(_MSC_VER)
   /*
@@ -1737,6 +1778,11 @@ Task_Switch(VMDriver *vm,  // IN
   uint32 pCPU;
   MPN hvRootMPN;
   Descriptor *tempGDTBase;
+#ifdef CYCLE_SPEC_CTRL
+   Bool specCtrlEqual = FALSE;
+   uint64 readSpecCtrlValue = 0;
+   static volatile uint64 currentSpecCtrlValue = 0;
+#endif

   ASSERT_ON_COMPILE(sizeof(VMCrossPage) == PAGE_SIZE);
   TaskDisableNMI(&vm->hostAPIC, &lint0NMI, &lint1NMI, &pcNMI, &thermalNMI);
@@ -1747,7 +1793,8 @@ Task_Switch(VMDriver *vm,  // IN
   ASSERT(pCPU < ARRAYSIZE(hvRootPage) && pCPU < ARRAYSIZE(tmpGDT));

   hvRootMPN = Atomic_Read64(&hvRootPage[pCPU]);
-   tempGDTBase = USE_TEMPORARY_GDT ? Atomic_ReadPtr(&tmpGDT[pCPU]) : NULL;
+   tempGDTBase = USE_TEMPORARY_GDT ? Atomic_ReadPtr(&tmpGDT[pCPU])
+                                   : default_rw_gdt;

   /*
    * We can't allocate memory with interrupts disabled on all hosts
@@ -1945,6 +1992,14 @@ Task_Switch(VMDriver *vm,  // IN
         TS_ASSERT(SELECTOR_TABLE(ds) == SELECTOR_GDT);
         TS_ASSERT(SELECTOR_TABLE(ss) == SELECTOR_GDT);

+         if (CPUID_HostSupportsSpecCtrl()) {
+#ifdef CYCLE_SPEC_CTRL
+            currentSpecCtrlValue = (currentSpecCtrlValue + 1) % 4;
+            __SET_MSR(MSR_SPEC_CTRL, currentSpecCtrlValue);
+#endif
+            crosspage->crosspageData.specCtrl = __GET_MSR(MSR_SPEC_CTRL);
+         }
+
         DEBUG_ONLY(crosspage->crosspageData.tinyStack[0] = 0xDEADBEEF;)
         /* Running in host context prior to TaskSwitchToMonitor() */
         TaskSwitchToMonitor(crosspage);
@@ -1952,6 +2007,16 @@ Task_Switch(VMDriver *vm,  // IN

         TS_ASSERT(crosspage->crosspageData.tinyStack[0] == 0xDEADBEEF);

+#ifdef CYCLE_SPEC_CTRL
+         if (CPUID_HostSupportsSpecCtrl()) {
+            readSpecCtrlValue = __GET_MSR(MSR_SPEC_CTRL);
+            specCtrlEqual = readSpecCtrlValue ==
+                            crosspage->crosspageData.specCtrl;
+            /* Do not leak cycling SPEC_CTRL value back to host. */
+            __SET_MSR(MSR_SPEC_CTRL, 0);
+         }
+#endif
+
         /*
          * Temporarily disable single-step stress as VMX/VMCS change code
          * ASSERTS on RFLAGS content without allowing TF/RF to be set.
@@ -2085,6 +2150,12 @@ Task_Switch(VMDriver *vm,  // IN
         if (UNLIKELY(TaskGotException(crosspage, EXC_UD))) {
            Warning("#UD occurred on switch back to host; dumping core");
         }
+#ifdef CYCLE_SPEC_CTRL
+         if (UNLIKELY(!specCtrlEqual)) {
+            Warning("SpecCtrl not equal: expected %"FMT64"x, got %"FMT64"x\n",
+                    crosspage->crosspageData.specCtrl, readSpecCtrlValue);
+         }
+#endif
         /*
          * The NMI/MCE checks above are special cases for interrupts
          * received during worldswitch.  Here is the more generic case
@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998-2016 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998-2017 VMware, Inc. All rights reserved.
 *
 * 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
@@ -720,6 +720,35 @@ cleanup:
 }


+/*
+ *----------------------------------------------------------------------
+ *
+ * Vmx86_LookupUserMPN --
+ *
+ *      Look up the MPN of a locked user page by user VA under the VM lock.
+ *
+ * Results:
+ *      A status code and the MPN on success.
+ *
+ * Side effects:
+ *      None
+ *
+ *----------------------------------------------------------------------
+ */
+
+int
+Vmx86_LookupUserMPN(VMDriver *vm, // IN: VMDriver
+                    VA64 uAddr,   // IN: user VA of the page
+                    MPN *mpn)     // OUT
+{
+   int ret;
+   HostIF_VMLock(vm, 38);
+   ret = HostIF_LookupUserMPN(vm, uAddr, mpn);
+   HostIF_VMUnlock(vm, 38);
+   return ret;
+}
+
+
 /*
 *----------------------------------------------------------------------
 *
@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998-2013,2015-2016 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998-2017 VMware, Inc. All rights reserved.
 *
 * 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
@@ -106,6 +106,7 @@ extern PseudoTSC pseudoTSC;
 #define MAX_LOCKED_PAGES (-1)

 extern VMDriver *Vmx86_CreateVM(void);
+extern int Vmx86_LookupUserMPN(VMDriver *vm, VA64 uAddr, MPN *mpn);
 extern int Vmx86_ReleaseVM(VMDriver *vm);
 extern int Vmx86_InitVM(VMDriver *vm, InitBlock *initParams);
 extern int Vmx86_LateInitVM(VMDriver *vm);
@@ -24,7 +24,11 @@
 * Include linux/cred.h via linux/sched.h - it is not nice, but
 * as cpp does not have #ifexist...
 */
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 11, 0)
 #include <linux/sched.h>
+#else
+#include <linux/cred.h>
+#endif

 #if !defined(current_fsuid) && LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 29)
 #define current_uid() (current->uid)
@@ -0,0 +1,34 @@
+#ifndef __COMPAT_MMAP_LOCK_H__
+#define __COMPAT_MMAP_LOCK_H__
+
+#include <linux/mm.h>
+
+/*
+ * In 5.8-rc1, mmap locking was reworked to use wrappers around mmap_sem
+ * (which was also renamed to mmap_lock). All code is now supposed to use
+ * these wrappers as the internal implementation of the lock may change in
+ * the future.
+ *
+ * Check also _LINUX_MMAP_LOCK_H to handle possible backports to distribution
+ * pre-5.8 kernel. This macro is defined in <linux/mmap_lock.h> which is also
+ * included in <linux/mm.h> since the commit introducing the wrappers so that
+ * we should have it defined in any kernel providing the new API.
+ */
+
+#if (LINUX_VERSION_CODE >= KERNEL_VERSION(5, 8, 0)) || defined(_LINUX_MMAP_LOCK_H)
+#include <linux/mmap_lock.h>
+#else
+
+static inline void mmap_read_lock(struct mm_struct *mm)
+{
+	down_read(&mm->mmap_sem);
+}
+
+static inline void mmap_read_unlock(struct mm_struct *mm)
+{
+	up_read(&mm->mmap_sem);
+}
+
+#endif /* 5.8.0 */
+
+#endif /* __COMPAT_MMAP_LOCK_H__ */
@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 2002 VMware, Inc. All rights reserved.
+ * Copyright (C) 2002-2017 VMware, Inc. All rights reserved.
 *
 * 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
@@ -30,80 +30,32 @@
 #include <asm/pgtable.h>


-/* pte_page() API modified in 2.3.23 to return a struct page * --hpreg */
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 3, 23)
-#   define compat_pte_page pte_page
-#else
-#   include "compat_page.h"
-
-#   define compat_pte_page(_pte) virt_to_page(pte_page(_pte))
-#endif
-
-
-/* Appeared in 2.5.5 --hpreg */
-#ifndef pte_offset_map
-/*  Appeared in SuSE 8.0's 2.4.18 --hpreg */
-#   ifdef pte_offset_atomic
-#      define pte_offset_map pte_offset_atomic
-#      define pte_unmap pte_kunmap
-#   else
-#      define pte_offset_map pte_offset
-#      define pte_unmap(_pte)
-#   endif
-#endif
-
-
-/* Appeared in 2.5.74-mmX --petr */
-#ifndef pmd_offset_map
-#   define pmd_offset_map(pgd, address) pmd_offset(pgd, address)
-#   define pmd_unmap(pmd)
-#endif
-
-
 /*
- * Appeared in 2.6.10-rc2-mm1.  Older kernels did L4 page tables as 
- * part of pgd_offset, or they did not have L4 page tables at all.
- * In 2.6.11 pml4 -> pgd -> pmd -> pte hierarchy was replaced by
- * pgd -> pud -> pmd -> pte hierarchy.
+ * p4d level appeared in 4.12.
 */
-#ifdef PUD_MASK
-#   define compat_pgd_offset(mm, address)   pgd_offset(mm, address)
-#   define compat_pgd_present(pgd)          pgd_present(pgd)
-#   define compat_pud_offset(pgd, address)  pud_offset(pgd, address)
-#   define compat_pud_present(pud)          pud_present(pud)
-typedef pgd_t  compat_pgd_t;
-typedef pud_t  compat_pud_t;
-#elif defined(pml4_offset)
-#   define compat_pgd_offset(mm, address)   pml4_offset(mm, address)
-#   define compat_pgd_present(pml4)         pml4_present(pml4)
-#   define compat_pud_offset(pml4, address) pml4_pgd_offset(pml4, address)
-#   define compat_pud_present(pgd)          pgd_present(pgd)
-typedef pml4_t compat_pgd_t;
-typedef pgd_t  compat_pud_t;
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 12, 0)
+#   define compat_p4d_offset(pgd, address)  p4d_offset(pgd, address)
+#   define compat_p4d_present(p4d)          p4d_present(p4d)
+#   define compat_p4d_large(p4d)            p4d_large(p4d)
+#   define compat_p4d_pfn(p4d)              p4d_pfn(p4d)
+#   define COMPAT_P4D_MASK                  P4D_MASK
+typedef p4d_t compat_p4d_t;
 #else
-#   define compat_pgd_offset(mm, address)   pgd_offset(mm, address)
-#   define compat_pgd_present(pgd)          pgd_present(pgd)
-#   define compat_pud_offset(pgd, address)  (pgd)
-#   define compat_pud_present(pud)          (1)
-typedef pgd_t  compat_pgd_t;
-typedef pgd_t  compat_pud_t;
+#   define compat_p4d_offset(pgd, address)  (pgd)
+#   define compat_p4d_present(p4d)          (1)
+#   define compat_p4d_large(p4d)            (0)
+#   define compat_p4d_pfn(p4d)              INVALID_MPN  /* Not used */
+#   define COMPAT_P4D_MASK                  0            /* Not used */
+typedef pgd_t compat_p4d_t;
 #endif
-
-
-#define compat_pgd_offset_k(mm, address) pgd_offset_k(address)
-
-
-/* Introduced somewhere in 2.6.0, + backported to some 2.4 RedHat kernels */
-#if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 0) && !defined(pte_pfn)
-#   define pte_pfn(pte) page_to_pfn(compat_pte_page(pte))
+/* p[gu]d_large did not exist before 2.6.25 */
+#if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 25)
+#   define pud_large(pud) 0
+#   define pgd_large(pgd) 0
 #endif
-
-
-/* A page_table_lock field is added to struct mm_struct in 2.3.10 --hpreg */
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(2, 3, 10)
-#   define compat_get_page_table_lock(_mm) (&(_mm)->page_table_lock)
-#else
-#   define compat_get_page_table_lock(_mm) NULL
+/* pud_pfn did not exist before 3.8. */
+#if LINUX_VERSION_CODE < KERNEL_VERSION(3, 8, 0)
+#   define pud_pfn(pud)  INVALID_MPN
 #endif


@@ -128,12 +80,8 @@ typedef pgd_t  compat_pud_t;
 #define VM_PAGE_KERNEL_EXEC PAGE_KERNEL
 #endif
 #else
-#ifdef PAGE_KERNEL_EXECUTABLE
-#define VM_PAGE_KERNEL_EXEC PAGE_KERNEL_EXECUTABLE
-#else
 #define VM_PAGE_KERNEL_EXEC PAGE_KERNEL_EXEC
 #endif
-#endif


 #endif /* __COMPAT_PGTABLE_H__ */
@@ -0,0 +1,30 @@
+#ifndef __COMPAT_POLL_H__
+#define __COMPAT_POLL_H__
+
+#include <linux/poll.h>
+
+#if (LINUX_VERSION_CODE < KERNEL_VERSION(4, 18, 0))
+
+#ifndef __poll_t
+typedef unsigned int __poll_t;
+#endif
+
+static inline __poll_t compat_vfs_poll(struct file *file,
+				       struct poll_table_struct *pt)
+{
+	if (unlikely(!file->f_op->poll))
+		return DEFAULT_POLLMASK;
+	return file->f_op->poll(file, pt);
+}
+
+#else
+
+static inline __poll_t compat_vfs_poll(struct file *file,
+				       struct poll_table_struct *pt)
+{
+	return vfs_poll(file, pt);
+}
+
+#endif
+
+#endif /* __COMPAT_POLL_H__ */
@@ -0,0 +1,33 @@
+#ifndef __COMPAT_TIMER_H__
+#   define __COMPAT_TIMER_H__
+
+#include <linux/timer.h>
+
+#if (LINUX_VERSION_CODE < KERNEL_VERSION(4, 15, 0)) && !defined(timer_setup)
+
+typedef unsigned long compat_timer_arg_t;
+
+static inline void compat_timer_setup(struct timer_list *timer,
+				      void (*func)(compat_timer_arg_t),
+				      unsigned int flags)
+{
+	init_timer(timer);
+	timer->function = func;
+	timer->data = 0;
+	timer->flags = flags;
+}
+
+#else /* new timer interface since 4.15 */
+
+typedef struct timer_list *compat_timer_arg_t;
+
+static inline void compat_timer_setup(struct timer_list *timer,
+				      void (*func)(compat_timer_arg_t),
+				      unsigned int flags)
+{
+	timer_setup(timer, func, flags);
+}
+
+#endif /* new timer interface since 4.15 */
+
+#endif /* __COMPAT_TIMER_H__ */
@@ -251,6 +251,7 @@ static INLINE ub8 hash2(register const ub8 *k, /* the key */
  {
    /* c is reserved for the length */
  case  2: b+=k[1];
+	   /* fall through */
  case  1: a+=k[0];
    /* case 0: nothing left to add */
  }
@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998-2015 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998-2015,2018 VMware, Inc. All rights reserved.
 *
 * 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
@@ -133,7 +133,7 @@ PtrToVA64(void const *ptr) // IN
 *
 */

-#define VMMON_VERSION           (308 << 16 | 0)
+#define VMMON_VERSION           (309 << 16 | 0)
 #define VMMON_VERSION_MAJOR(v)  ((uint32) (v) >> 16)
 #define VMMON_VERSION_MINOR(v)  ((uint16) (v))

@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998-2015 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998-2015,2018 VMware, Inc. All rights reserved.
 *
 * 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
@@ -421,6 +421,7 @@ struct VMCrossPageData {
    */
   uint8 switchHostIDT [sizeof(Gate64) * NUM_EXCEPTIONS]; // hostCS:hostVA
   uint8 switchMon64IDT[sizeof(Gate64) * NUM_EXCEPTIONS]; // 64-bit monCS:monVA
+   uint64        specCtrl; /* host MSR_SPEC_CTRL value before world switch. */
 }
 #include "vmware_pack_end.h"
 VMCrossPageData;
@@ -474,7 +475,7 @@ struct VMCrossPage {
 #include "vmware_pack_end.h"
 VMCrossPage;

-#define CROSSPAGE_VERSION_BASE 0xbf1 /* increment by 1 */
+#define CROSSPAGE_VERSION_BASE 0xbf2 /* increment by 1 */
 #define CROSSPAGE_VERSION    ((CROSSPAGE_VERSION_BASE << 1) + WS_INTR_STRESS)

 #if !defined(VMX86_SERVER) && defined(VMM)
@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 2002,2014 VMware, Inc. All rights reserved.
+ * Copyright (C) 2002,2014-2017 VMware, Inc. All rights reserved.
 *
 * 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
@@ -26,154 +26,14 @@
 #include "compat_spinlock.h"
 #include "compat_page.h"

-/*
- *-----------------------------------------------------------------------------
- *
- * PgtblPte2MPN --
- *
- *    Returns the page structure associated to a Page Table Entry.
- *
- *    This function is not allowed to schedule() because it can be called while
- *    holding a spinlock --hpreg
- *
- * Results:
- *    INVALID_MPN on failure
- *    mpn         on success
- *
- * Side effects:
- *    None
- *
- *-----------------------------------------------------------------------------
- */
-
-static INLINE MPN
-PgtblPte2MPN(pte_t *pte)   // IN
-{
-   MPN mpn;
-   if (pte_present(*pte) == 0) {
-      return INVALID_MPN;
-   }
-   mpn = pte_pfn(*pte);
-   if (mpn >= INVALID_MPN) {
-      return INVALID_MPN;
-   }
-   return mpn;
-}
-
-
-/*
- *-----------------------------------------------------------------------------
- *
- * PgtblPte2Page --
- *
- *    Returns the page structure associated to a Page Table Entry.
- *
- *    This function is not allowed to schedule() because it can be called while
- *    holding a spinlock --hpreg
- *
- * Results:
- *    The page structure if the page table entry points to a physical page
- *    NULL if the page table entry does not point to a physical page
- *
- * Side effects:
- *    None
- *
- *-----------------------------------------------------------------------------
- */
-
-static INLINE struct page *
-PgtblPte2Page(pte_t *pte) // IN
-{
-   if (pte_present(*pte) == 0) {
-      return NULL;
-   }
-
-   return compat_pte_page(*pte);
-}
-
-
-/*
- *-----------------------------------------------------------------------------
- *
- * PgtblPGD2PTELocked --
- *
- *    Walks through the hardware page tables to try to find the pte
- *    associated to a virtual address.
- *
- * Results:
- *    pte. Caller must call pte_unmap if valid pte returned.
- *
- * Side effects:
- *    None
- *
- *-----------------------------------------------------------------------------
- */
-
-static INLINE pte_t *
-PgtblPGD2PTELocked(compat_pgd_t *pgd,    // IN: PGD to start with
-                   VA addr)              // IN: Address in the virtual address
-                                         //     space of that process
-{
-   compat_pud_t *pud;
-   pmd_t *pmd;
-   pte_t *pte;
-
-   if (compat_pgd_present(*pgd) == 0) {
-      return NULL;
-   }
-
-   pud = compat_pud_offset(pgd, addr);
-   if (compat_pud_present(*pud) == 0) {
-      return NULL;
-   }
-
-   pmd = pmd_offset_map(pud, addr);
-   if (pmd_present(*pmd) == 0) {
-      pmd_unmap(pmd);
-      return NULL;
-   }
-
-   pte = pte_offset_map(pmd, addr);
-   pmd_unmap(pmd);
-   return pte;
-}
-
-
-/*
- *-----------------------------------------------------------------------------
- *
- * PgtblVa2PTELocked --
- *
- *    Walks through the hardware page tables to try to find the pte
- *    associated to a virtual address.
- *
- * Results:
- *    pte. Caller must call pte_unmap if valid pte returned.
- *
- * Side effects:
- *    None
- *
- *-----------------------------------------------------------------------------
- */
-
-static INLINE pte_t *
-PgtblVa2PTELocked(struct mm_struct *mm, // IN: Mm structure of a process
-                  VA addr)              // IN: Address in the virtual address
-                                        //     space of that process
-{
-   return PgtblPGD2PTELocked(compat_pgd_offset(mm, addr), addr);
-}
-

 /*
 *-----------------------------------------------------------------------------
 *
 * PgtblVa2MPNLocked --
 *
- *    Retrieve MPN for a given va.
- *
- *    Caller must call pte_unmap if valid pte returned. The mm->page_table_lock
- *    must be held, so this function is not allowed to schedule() --hpreg
+ *    Walks through the hardware page tables to try to find the pte
+ *    associated to a virtual address.  Then maps PTE to MPN.
 *
 * Results:
 *    INVALID_MPN on failure
@@ -188,89 +48,64 @@ PgtblVa2PTELocked(struct mm_struct *mm, // IN: Mm structure of a process
 static INLINE MPN
 PgtblVa2MPNLocked(struct mm_struct *mm, // IN: Mm structure of a process
                  VA addr)              // IN: Address in the virtual address
+                                        //     space of that process
 {
-   pte_t *pte;
+   pgd_t *pgd;
+   compat_p4d_t *p4d;
+   MPN mpn;

-   pte = PgtblVa2PTELocked(mm, addr);
-   if (pte != NULL) {
-      MPN mpn = PgtblPte2MPN(pte);
-      pte_unmap(pte);
-      return mpn;
+   pgd = pgd_offset(mm, addr);
+   if (pgd_present(*pgd) == 0) {
+      return INVALID_MPN;
   }
-   return INVALID_MPN;
-}
-
-
-#if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 0)
-/*
- *-----------------------------------------------------------------------------
- *
- * PgtblKVa2MPNLocked --
- *
- *    Retrieve MPN for a given kernel va.
- *
- *    Caller must call pte_unmap if valid pte returned. The mm->page_table_lock
- *    must be held, so this function is not allowed to schedule() --hpreg
- *
- * Results:
- *    INVALID_MPN on failure
- *    mpn         on success
- *
- * Side effects:
- *    None
- *
- *-----------------------------------------------------------------------------
- */
-
-static INLINE MPN
-PgtblKVa2MPNLocked(struct mm_struct *mm, // IN: Mm structure of a caller
-                   VA addr)              // IN: Address in the virtual address
-{
-   pte_t *pte;
-
-   pte = PgtblPGD2PTELocked(compat_pgd_offset_k(mm, addr), addr);
-   if (pte != NULL) {
-      MPN mpn = PgtblPte2MPN(pte);
-      pte_unmap(pte);
-      return mpn;
+   if (pgd_large(*pgd)) {
+      /* Linux kernel does not support PGD huge pages. */
+      /* return pgd_pfn(*pgd) + ((addr & PGD_MASK) >> PAGE_SHIFT); */
+      return INVALID_MPN;
   }
-   return INVALID_MPN;
-}
-#endif

-
-/*
- *-----------------------------------------------------------------------------
- *
- * PgtblVa2PageLocked --
- *
- *    Return the "page" struct for a given va.
- *
- * Results:
- *    struct page or NULL.  The mm->page_table_lock must be held, so this 
- *    function is not allowed to schedule() --hpreg
- *
- * Side effects:
- *    None
- *
- *-----------------------------------------------------------------------------
- */
-
-static INLINE struct page *
-PgtblVa2PageLocked(struct mm_struct *mm, // IN: Mm structure of a process
-                   VA addr)              // IN: Address in the virtual address
-{
-   pte_t *pte;
-
-   pte = PgtblVa2PTELocked(mm, addr);
-   if (pte != NULL) {
-      struct page *page = PgtblPte2Page(pte);
-      pte_unmap(pte);
-      return page;
+   p4d = compat_p4d_offset(pgd, addr);
+   if (compat_p4d_present(*p4d) == 0) {
+      return INVALID_MPN;
+   }
+   if (compat_p4d_large(*p4d)) {
+      mpn = compat_p4d_pfn(*p4d) + ((addr & ~COMPAT_P4D_MASK) >> PAGE_SHIFT);
   } else {
-      return NULL;
+      pud_t *pud;
+
+      pud = pud_offset(p4d, addr);
+      if (pud_present(*pud) == 0) {
+         return INVALID_MPN;
+      }
+      if (pud_large(*pud)) {
+         mpn = pud_pfn(*pud) + ((addr & ~PUD_MASK) >> PAGE_SHIFT);
+      } else {
+         pmd_t *pmd;
+
+         pmd = pmd_offset(pud, addr);
+         if (pmd_present(*pmd) == 0) {
+            return INVALID_MPN;
+         }
+         if (pmd_large(*pmd)) {
+            mpn = pmd_pfn(*pmd) + ((addr & ~PMD_MASK) >> PAGE_SHIFT);
+         } else {
+            pte_t *pte;
+
+            pte = pte_offset_map(pmd, addr);
+            if (pte_present(*pte) == 0) {
+               pte_unmap(pte);
+               return INVALID_MPN;
+            }
+            mpn = pte_pfn(*pte);
+            pte_unmap(pte);
+         }
+      }
   }
-} 
+   if (mpn >= INVALID_MPN) {
+      mpn = INVALID_MPN;
+   }
+   return mpn;
+}


 /*
@@ -298,85 +133,10 @@ PgtblVa2MPN(VA addr)  // IN

   /* current->mm is NULL for kernel threads, so use active_mm. */
   mm = current->active_mm;
-   if (compat_get_page_table_lock(mm)) {
-      spin_lock(compat_get_page_table_lock(mm));
-   }
+   spin_lock(&mm->page_table_lock);
   mpn = PgtblVa2MPNLocked(mm, addr);
-   if (compat_get_page_table_lock(mm)) {
-      spin_unlock(compat_get_page_table_lock(mm));
-   }
+   spin_unlock(&mm->page_table_lock);
   return mpn;
 }

-
-#if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 0)
-/*
- *-----------------------------------------------------------------------------
- *
- * PgtblKVa2MPN --
- *
- *    Walks through the hardware page tables of the current process to try to
- *    find the page structure associated to a virtual address.
- *
- * Results:
- *    Same as PgtblVa2MPNLocked()
- *
- * Side effects:
- *    None
- *
- *-----------------------------------------------------------------------------
- */
-
-static INLINE MPN
-PgtblKVa2MPN(VA addr)  // IN
-{
-   struct mm_struct *mm = current->active_mm;
-   MPN mpn;
-
-   if (compat_get_page_table_lock(mm)) {
-      spin_lock(compat_get_page_table_lock(mm));
-   }
-   mpn = PgtblKVa2MPNLocked(mm, addr);
-   if (compat_get_page_table_lock(mm)) {
-      spin_unlock(compat_get_page_table_lock(mm));
-   }
-   return mpn;
-}
-#endif
-
-
-/*
- *-----------------------------------------------------------------------------
- *
- * PgtblVa2Page --
- *
- *    Walks through the hardware page tables of the current process to try to
- *    find the page structure associated to a virtual address.
- *
- * Results:
- *    Same as PgtblVa2PageLocked()
- *
- * Side effects:
- *    None
- *
- *-----------------------------------------------------------------------------
- */
-
-static INLINE struct page *
-PgtblVa2Page(VA addr) // IN
-{
-   struct mm_struct *mm = current->active_mm;
-   struct page *page;
-
-   if (compat_get_page_table_lock(mm)) {
-      spin_lock(compat_get_page_table_lock(mm));
-   }
-   page = PgtblVa2PageLocked(mm, addr);
-   if (compat_get_page_table_lock(mm)) {
-      spin_unlock(compat_get_page_table_lock(mm));
-   }
-   return page;
-}
-
-
 #endif /* __PGTBL_H__ */
@@ -67,6 +67,7 @@ extern "C" {
 #if defined (VMKPANIC) 
 #include "vmk_assert.h"
 #else /* !VMKPANIC */
+#include <linux/kernel.h>
 #define _ASSERT_PANIC(name) \
           Panic(_##name##Fmt "\n", __FILE__, __LINE__)
 #define _ASSERT_PANIC_BUG(bug, name) \
@@ -118,7 +119,7 @@ NORETURN void Panic_NoSaveNoReturn(const char *fmt, ...) PRINTF_DECL(1, 2);
 } while(0)

 #else
-NORETURN void Panic(const char *fmt, ...) PRINTF_DECL(1, 2);
+#define Panic panic
 #endif

 void LogThrottled(uint32 *count, const char *fmt, ...) PRINTF_DECL(2, 3);
@@ -35,6 +35,8 @@
 #define INCLUDE_ALLOW_VMCORE
 #include "includeCheck.h"

+#include <asm/processor-flags.h>
+
 #define X86_MAX_INSTR_LEN  15   /* Max byte length of an x86 instruction. */

 #define NUM_IDT_VECTORS 256
@@ -62,7 +64,9 @@
 #define CR3_PDB_MASK   0xfffff000
 #define CR3_IGNORE     0xFFF
 #define PAE_CR3_IGNORE 0x1F
+#ifndef CR3_PCID_MASK
 #define CR3_PCID_MASK  0xFFF
+#endif
 #define CR3_NO_FLUSH   (1ULL << 63)

 #define CR4_VME        0x00000001
@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998-2016 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998-2017 VMware, Inc. All rights reserved.
 *
 * 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
@@ -419,6 +419,10 @@ FIELD(  6,  0, EBX,  0,  4, NUM_INTR_THRESHOLDS,                   NA,  FALSE) \
 FLAG(   6,  0, ECX,  0,  1, HW_COORD_FEEDBACK,                     NA,  FALSE) \
 FLAG(   6,  0, ECX,  3,  1, ENERGY_PERF_BIAS,                      NA,  FALSE)

+#define CPUID_7_EDX_26 \
+FLAG(   7,  0, EDX, 26,  1, LEVEL7EDX_RSVD1,                       NO,  FALSE)
+#define CPUID_7_EDX_27 \
+FLAG(   7,  0, EDX, 27,  1, LEVEL7EDX_RSVD2,                       NO,  FALSE)

 /*    LEVEL, SUB-LEVEL, REG, POS, SIZE, NAME,                  MON SUPP, CPL3 */
 #define CPUID_FIELD_DATA_LEVEL_7                                               \
@@ -440,8 +444,9 @@ FLAG(   7,  0, EBX, 18,  1, RDSEED,                                YES, TRUE)  \
 FLAG(   7,  0, EBX, 19,  1, ADX,                                   YES, TRUE)  \
 FLAG(   7,  0, EBX, 20,  1, SMAP,                                  YES, FALSE) \
 FLAG(   7,  0, EBX, 25,  1, PT,                                    NO,  FALSE) \
-FLAG(   7,  0, ECX,  0,  1, PREFETCHWT1,                           NO,  TRUE)
-
+FLAG(   7,  0, ECX,  0,  1, PREFETCHWT1,                           NO,  TRUE)  \
+CPUID_7_EDX_26 \
+CPUID_7_EDX_27

 /*    LEVEL, SUB-LEVEL, REG, POS, SIZE, NAME,                  MON SUPP, CPL3 */
 #define CPUID_FIELD_DATA_LEVEL_A                                               \
@@ -708,6 +713,7 @@ FIELD( 84,  0, EBX,  0, 32, LEAF84_BRAND_STRING_EBX,               NA,  FALSE) \
 FIELD( 84,  0, ECX,  0, 32, LEAF84_BRAND_STRING_ECX,               NA,  FALSE) \
 FIELD( 84,  0, EDX,  0, 32, LEAF84_BRAND_STRING_EDX,               NA,  FALSE)

+
 #define CPUID_8A_EDX_11 \
 FLAG(  8A,  0, EDX, 11,  1, SVMEDX_RSVD1,                          NO,  FALSE)
 #define CPUID_8A_EDX_14_31 \
@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 2003-2015 VMware, Inc. All rights reserved.
+ * Copyright (C) 2003-2015,2018 VMware, Inc. All rights reserved.
 *
 * 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
@@ -63,6 +63,14 @@ extern "C" {
 */
 void __cpuid(int regs[4], int eax);
 #pragma intrinsic(__cpuid)
+
+/*
+ * __cpuidex has been supported since VS2008
+ */
+#if _MSC_VER >= 1500
+void __cpuidex(int regs[4], int eax, int ecx);
+#pragma intrinsic(__cpuidex)
+#endif /* _MSC_VER >= 1500 */
 #endif /* VM_X86_64 */

 #ifdef __cplusplus
@@ -247,6 +255,20 @@ __GET_CPUID(int input, CPUIDRegs *regs)

 #ifdef VM_X86_64

+#if _MSC_VER >= 1500
+
+/*
+ * __cpuidex has been supported since VS2008
+ */
+
+static INLINE void
+__GET_CPUID2(int inputEax, int inputEcx, CPUIDRegs *regs)
+{
+   __cpuidex((int *)regs, inputEax, inputEcx);
+}
+
+#else // _MSC_VER >= 1500
+
 /*
 * No inline assembly in Win64. Implemented in bora/lib/misc in
 * cpuidMasm64.asm.
@@ -254,6 +276,7 @@ __GET_CPUID(int input, CPUIDRegs *regs)

 extern void
 __GET_CPUID2(int inputEax, int inputEcx, CPUIDRegs *regs);
+#endif // _MSC_VER >= 1500

 #else // VM_X86_64

@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998-2015 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998-2018 VMware, Inc. All rights reserved.
 *
 * 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
@@ -24,6 +24,7 @@

 #ifndef _X86MSR_H_
 #define _X86MSR_H_
+#include <asm/msr-index.h>
 #define INCLUDE_ALLOW_USERLEVEL
 #define INCLUDE_ALLOW_VMX

@@ -79,6 +80,7 @@ MSRQuery;
 #define MSR_SMI_COUNT         0x00000034 // Intel Nehalem Family
 #define MSR_FEATCTL           0x0000003a
 #define MSR_TSC_ADJUST        0x0000003b
+#define MSR_SPEC_CTRL         0x00000048
 #define MSR_BIOS_UPDT_TRIG    0x00000079
 #define MSR_BIOS_SIGN_ID      0x0000008b
 #define MSR_PERFCTR0          0x000000c1
@@ -106,7 +108,10 @@ MSRQuery;
 #define MSR_TSC_AUX           0xc0000103
 #define MSR_BD_TSC_RATIO      0xc0000104

+
+#ifndef MSR_MISC_FEATURES_ENABLES
 #define MSR_MISC_FEATURES_ENABLES            0x140
+#endif

 /* Intel Core Architecture and later: use only architected counters. */
 #define IA32_MSR_PERF_CAPABILITIES                0x345
@@ -300,7 +305,9 @@ typedef enum {
 #define MSR_K7_HWCR_SSEDIS         0x00008000ULL // Disable SSE bit
 #define MSR_K7_HWCR_MONMWAITUSEREN 0x00000400ULL // Enable MONITOR/MWAIT CPL>0
 #define MSR_K7_HWCR_TLBFFDIS       0x00000040ULL // Disable TLB Flush Filter
+#ifndef MSR_K7_HWCR_SMMLOCK
 #define MSR_K7_HWCR_SMMLOCK        0x00000001ULL // Lock SMM environment
+#endif

 #ifndef MSR_K8_SYSCFG
 #define MSR_K8_SYSCFG        0xc0010010
@@ -342,6 +349,8 @@ typedef enum {
 #define MSR_VM_CR_SVME_DISABLE     0x0000000000000010ULL
 #define MSR_VM_CR_RESERVED         0xffffffffffffffe0ULL

+#define MSR_AMD_DE_CFG           0xc0011029  // Decode configuration
+#define MSR_AMD_DE_CFG_BIT1      (1ULL<<1)

 /* Syscall/Sysret related MSRs (x86_64) */
 #define MSR_STAR             0xc0000081 // Also present on Athlons.
@@ -475,7 +484,11 @@ typedef enum {
 /*
 * MISC_FEATURES_ENABLES bits
 */
+#ifdef MSR_MISC_FEATURES_ENABLES_CPUID_FAULT
+#define MSR_MISC_FEATURES_ENABLES_CPUID_FAULTING MSR_MISC_FEATURES_ENABLES_CPUID_FAULT
+#else
 #define MSR_MISC_FEATURES_ENABLES_CPUID_FAULTING 1
+#endif



@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998-2016 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998-2017 VMware, Inc. All rights reserved.
 *
 * 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
@@ -21,6 +21,7 @@

 #define EXPORT_SYMTAB

+#include "compat_timer.h"
 #include <linux/file.h>
 #include <linux/highmem.h>
 #include <linux/poll.h>
@@ -80,6 +81,9 @@ static Bool LinuxDriverCheckPadding(void);

 struct VMXLinuxState linuxState;

+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 17, 0)
+typedef int vm_fault_t;
+#endif

 /*
 *----------------------------------------------------------------------
@@ -104,8 +108,10 @@ long LinuxDriver_Ioctl(struct file *filp, u_int iocmd,

 static int LinuxDriver_Close(struct inode *inode, struct file *filp);
 static unsigned int LinuxDriverPoll(struct file *file, poll_table *wait);
-#if defined(VMW_NOPAGE_2624)
-static int LinuxDriverFault(struct vm_area_struct *vma, struct vm_fault *fault);
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 11, 0)
+static vm_fault_t LinuxDriverFault(struct vm_fault *fault);
+#elif defined(VMW_NOPAGE_2624)
+static vm_fault_t LinuxDriverFault(struct vm_area_struct *vma, struct vm_fault *fault);
 #else
 static struct page *LinuxDriverNoPage(struct vm_area_struct *vma,
                                      unsigned long address,
@@ -113,7 +119,7 @@ static struct page *LinuxDriverNoPage(struct vm_area_struct *vma,
 #endif
 static int LinuxDriverMmap(struct file *filp, struct vm_area_struct *vma);

-static void LinuxDriverPollTimeout(unsigned long clientData);
+static void LinuxDriverPollTimeout(compat_timer_arg_t unused);
 static unsigned int LinuxDriverEstimateTSCkHz(void);

 static struct vm_operations_struct vmuser_mops = {
@@ -225,7 +231,7 @@ LinuxDriverEstimateTSCkHz(void)
 *----------------------------------------------------------------------
 */
 static void
-LinuxDriverEstimateTSCkHzDeferred(unsigned long data)
+LinuxDriverEstimateTSCkHzDeferred(compat_timer_arg_t unused)
 {
   LinuxDriverEstimateTSCkHz();
 }
@@ -263,9 +269,7 @@ LinuxDriverInitTSCkHz(void)
   }

   Vmx86_ReadTSCAndUptime(&tsckHzStartTime);
-   tscTimer.function = LinuxDriverEstimateTSCkHzDeferred;
   tscTimer.expires  = jiffies + 4 * HZ;
-   tscTimer.data     = 0;
   add_timer(&tscTimer);
 }

@@ -307,9 +311,7 @@ init_module(void)
    */

   init_waitqueue_head(&linuxState.pollQueue);
-   init_timer(&linuxState.pollTimer);
-   linuxState.pollTimer.data = 0;
-   linuxState.pollTimer.function = LinuxDriverPollTimeout;
+   compat_timer_setup(&linuxState.pollTimer, LinuxDriverPollTimeout, 0);

   linuxState.fastClockThread = NULL;
   linuxState.fastClockFile = NULL;
@@ -358,7 +360,7 @@ init_module(void)
       linuxState.deviceName, linuxState.major, linuxState.minor);

   HostIF_InitUptime();
-   init_timer(&tscTimer);
+   compat_timer_setup(&tscTimer, LinuxDriverEstimateTSCkHzDeferred, 0);
   LinuxDriverInitTSCkHz();
   Vmx86_InitIDList();

@@ -739,14 +741,12 @@ void
 LinuxDriverWakeUp(Bool selective)  // IN:
 {
   if (selective && linuxState.pollList != NULL) {
-      struct timeval tv;
      VmTimeType now;
      VMLinux *p;
      VMLinux *next;

      HostIF_PollListLock(1);
-      do_gettimeofday(&tv);
-      now = tv.tv_sec * 1000000ULL + tv.tv_usec;
+      now = ktime_get_ns() / NSEC_PER_USEC;

      for (p = linuxState.pollList; p != NULL; p = next) {
         next = p->pollForw;
@@ -813,12 +813,10 @@ LinuxDriverPoll(struct file *filp,  // IN:
      }
   } else {
      if (linuxState.fastClockThread && vmLinux->pollTimeoutPtr != NULL) {
-         struct timeval tv;
+         u64 now = ktime_get_ns() / NSEC_PER_USEC;

-         do_gettimeofday(&tv);
         poll_wait(filp, &vmLinux->pollQueue, wait);
-         vmLinux->pollTime = *vmLinux->pollTimeoutPtr +
-                                       tv.tv_sec * 1000000ULL + tv.tv_usec;
+         vmLinux->pollTime = *vmLinux->pollTimeoutPtr + now;
         if (vmLinux->pollBack == NULL) {
            HostIF_PollListLock(2);
            if (vmLinux->pollBack == NULL) {
@@ -856,7 +854,7 @@ LinuxDriverPoll(struct file *filp,  // IN:
 */

 static void
-LinuxDriverPollTimeout(unsigned long clientData)  // IN:
+LinuxDriverPollTimeout(compat_timer_arg_t unused)  // IN:
 {
   LinuxDriverWakeUp(FALSE);
 }
@@ -881,15 +879,21 @@ LinuxDriverPollTimeout(unsigned long clientData)  // IN:
 *-----------------------------------------------------------------------------
 */

-#if defined(VMW_NOPAGE_2624)
-static int LinuxDriverFault(struct vm_area_struct *vma, //IN
-                            struct vm_fault *fault)     //IN/OUT
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 11, 0)
+static vm_fault_t
+LinuxDriverFault(struct vm_fault *fault)     //IN/OUT
+#elif defined(VMW_NOPAGE_2624)
+static vm_fault_t LinuxDriverFault(struct vm_area_struct *vma, //IN
+				   struct vm_fault *fault)     //IN/OUT
 #else
 static struct page *LinuxDriverNoPage(struct vm_area_struct *vma, //IN
                                      unsigned long address,      //IN
                                      int *type)                  //OUT: Fault type
 #endif
 {
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 11, 0)
+   struct vm_area_struct *vma = fault->vma;
+#endif
   VMLinux *vmLinux = (VMLinux *) vma->vm_file->private_data;
   unsigned long pg;
   struct page* page;
@@ -1279,7 +1283,7 @@ LinuxDriverReadTSC(void *data,   // OUT: TSC values
 *-----------------------------------------------------------------------------
 */

-__attribute__((always_inline)) static Bool
+__always_inline static Bool
 LinuxDriverSyncReadTSCs(uint64 *delta) // OUT: TSC max - TSC min
 {
   TSCDelta tscDelta;
@@ -1350,7 +1354,6 @@ LinuxDriver_Ioctl(struct file *filp,    // IN:
   case IOCTL_VMX86_CREATE_VM:
   case IOCTL_VMX86_INIT_CROSSGDT:
   case IOCTL_VMX86_SET_UID:
-   case IOCTL_VMX86_LOOK_UP_MPN:
   case IOCTL_VMX86_GET_NUM_VMS:
   case IOCTL_VMX86_GET_TOTAL_MEM_USAGE:
   case IOCTL_VMX86_SET_HARD_LIMIT:
@@ -1360,8 +1363,6 @@ LinuxDriver_Ioctl(struct file *filp,    // IN:
   case IOCTL_VMX86_GET_KHZ_ESTIMATE:
   case IOCTL_VMX86_GET_ALL_CPUID:
   case IOCTL_VMX86_GET_ALL_MSRS:
-   case IOCTL_VMX86_READ_PAGE:
-   case IOCTL_VMX86_WRITE_PAGE:
   case IOCTL_VMX86_SET_POLL_TIMEOUT_PTR:
   case IOCTL_VMX86_GET_KERNEL_CLOCK_RATE:
   case IOCTL_VMX86_GET_REFERENCE_CLOCK_HZ:
@@ -1522,7 +1523,7 @@ LinuxDriver_Ioctl(struct file *filp,    // IN:
      if (retval) {
         break;
      }
-      args.ret.status = HostIF_LookupUserMPN(vm, args.uAddr, &args.ret.mpn);
+      args.ret.status = Vmx86_LookupUserMPN(vm, args.uAddr, &args.ret.mpn);
      retval = HostIF_CopyToUser((void *)ioarg, &args, sizeof args);
      break;
   }
@@ -1841,7 +1842,7 @@ LinuxDriver_Ioctl(struct file *filp,    // IN:
         if (retval) {
            break;
         }
-         retval = HostIF_ReadPage(req.mpn, req.uAddr, FALSE);
+         retval = HostIF_ReadPage(vm, req.mpn, req.uAddr, FALSE);
         break;
      }

@@ -1852,7 +1853,7 @@ LinuxDriver_Ioctl(struct file *filp,    // IN:
         if (retval) {
            break;
         }
-         retval = HostIF_WritePage(req.mpn, req.uAddr, FALSE);
+         retval = HostIF_WritePage(vm, req.mpn, req.uAddr, FALSE);
         break;
      }

@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998-2014 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998-2017 VMware, Inc. All rights reserved.
 *
 * 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
@@ -30,6 +30,7 @@

 /* Must come before vmware.h --hpreg */
 #include "compat_page.h"
+#include "compat_timer.h"
 #include <linux/binfmts.h>
 #include <linux/delay.h>
 #include <linux/file.h>
@@ -68,24 +69,31 @@
 #endif

 #include <asm/io.h>
+#include <asm/page.h>
+#include <asm/tlbflush.h>
 #include <asm/uaccess.h>
+#include <asm/irq_vectors.h>
 #include <linux/mc146818rtc.h>
 #include <linux/capability.h>
 #include <linux/kthread.h>
 #include <linux/wait.h>
-
+#include <linux/signal.h>
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 11, 0)
+#include <linux/taskstats_kern.h> // For linux/sched/signal.h without version check
+#endif
+#include <linux/eventfd.h>

 #include "vmware.h"
 #include "x86apic.h"
 #include "vm_asm.h"
 #include "modulecall.h"
+#include "driver.h"
 #include "memtrack.h"
 #include "phystrack.h"
 #include "cpuid.h"
 #include "cpuid_info.h"
 #include "hostif.h"
 #include "hostif_priv.h"
-#include "driver.h"
 #include "vmhost.h"
 #include "x86msr.h"
 #include "apic.h"
@@ -95,6 +103,41 @@
 #include "pgtbl.h"
 #include "vmmonInt.h"
 #include "versioned_atomic.h"
+#include "compat_poll.h"
+#include "compat_mmap_lock.h"
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 14, 0)
+#   define global_zone_page_state global_page_state
+#endif
+
+static unsigned long get_nr_slab_unreclaimable(void)
+{
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 9, 0)
+   return global_node_page_state_pages(NR_SLAB_UNRECLAIMABLE_B);
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(4, 13, 0)
+   return global_node_page_state(NR_SLAB_UNRECLAIMABLE);
+#else
+   return global_page_state(NR_SLAB_UNRECLAIMABLE);
+#endif
+}
+
+static unsigned long get_nr_unevictable(void)
+{
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 8, 0)
+   return global_node_page_state(NR_UNEVICTABLE);
+#else
+   return global_page_state(NR_UNEVICTABLE);
+#endif
+}
+
+static unsigned long get_nr_anon_mapped(void)
+{
+ #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 8, 0)
+   return global_node_page_state(NR_ANON_MAPPED);
+ #else
+   return global_page_state(NR_ANON_PAGES);
+ #endif
+}

 /*
 * Determine if we can use high resolution timers.
@@ -139,13 +182,7 @@ static DECLARE_TASKLET(timerTasklet, HostIFWakeupClockThread, 0);
 */
 #define LOCKED_PAGE_SLACK 10000

-static struct {
-   Atomic_uint64     uptimeBase;
-   VersionedAtomic   version;
-   uint64            monotimeBase;
-   unsigned long     jiffiesBase;
-   struct timer_list timer;
-} uptimeState;
+u64 uptime_base;

 /*
 * First Page Locking strategy
@@ -204,6 +241,21 @@ static struct {
 uint8 monitorIPIVector;
 uint8 hvIPIVector;

+static unsigned long compat_totalram_pages(void)
+{
+#if LINUX_VERSION_CODE < KERNEL_VERSION(5, 0, 0)
+	return totalram_pages;
+#else
+	return totalram_pages();
+#endif
+}
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(5, 0, 0) && defined(VERIFY_WRITE)
+	#define write_access_ok(addr, size) access_ok(VERIFY_WRITE, addr, size)
+#else
+	#define write_access_ok(addr, size) access_ok(addr, size)
+#endif
+
 /*
 *-----------------------------------------------------------------------------
 *
@@ -626,6 +678,15 @@ HostIF_FastClockUnlock(int callerID) // IN
   MutexUnlock(&fastClockMutex, callerID);
 }

+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 8, 0)
+static int crosspage_set_exec(pte_t *pte, unsigned long addr, void *data)
+{
+	struct page *p = data;
+
+	set_pte(pte, mk_pte(p, VM_PAGE_KERNEL_EXEC));
+	return 0;
+}
+#endif

 /*
 *-----------------------------------------------------------------------------
@@ -692,7 +753,29 @@ HostIF_PollListUnlock(int callerID) // IN
 static void *
 MapCrossPage(struct page *p)  // IN:
 {
+#if COMPAT_LINUX_VERSION_CHECK_LT(5, 8, 0)
   return vmap(&p, 1, VM_MAP, VM_PAGE_KERNEL_EXEC);
+#else
+   void *addr;
+
+   addr = vmap(&p, 1, VM_MAP, VM_PAGE_KERNEL_EXEC);
+   if (!addr)
+	   return NULL;
+
+   /* Starting with 5.8, vmap() always sets the NX bit, but the cross
+    * page needs to be executable. */
+   if (apply_to_page_range(current->mm, (unsigned long)addr, PAGE_SIZE,
+			   crosspage_set_exec, p)) {
+      vunmap(addr);
+      return NULL;
+   }
+
+   preempt_disable();
+   __flush_tlb_all();
+   preempt_enable();
+
+   return addr;
+#endif
 }


@@ -1006,7 +1089,7 @@ HostIF_FreeLockedPages(VMDriver *vm,	     // IN: VM instance pointer
 int
 HostIF_Init(VMDriver *vm)  // IN:
 {
-   vm->memtracker = MemTrack_Init();
+   vm->memtracker = MemTrack_Init(vm);
   if (vm->memtracker == NULL) {
      return -1;
   }
@@ -1161,14 +1244,16 @@ HostIFGetUserPages(void *uvAddr,          // IN
 {
   int retval;

-   down_read(&current->mm->mmap_sem);
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 6, 0)
+   mmap_read_lock(current->mm);
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 9, 0)
+   retval = get_user_pages((unsigned long)uvAddr, numPages, 0, ppages, NULL);
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(4, 6, 0)
   retval = get_user_pages((unsigned long)uvAddr, numPages, 0, 0, ppages, NULL);
 #else
   retval = get_user_pages(current, current->mm, (unsigned long)uvAddr,
                           numPages, 0, 0, ppages, NULL);
 #endif
-   up_read(&current->mm->mmap_sem);
+   mmap_read_unlock(current->mm);

   return retval != numPages;
 }
@@ -1564,14 +1649,7 @@ unsigned int
 HostIF_EstimateLockedPageLimit(const VMDriver* vm,                // IN
 			       unsigned int currentlyLockedPages) // IN
 {
-   /*
-    * This variable is available and exported to modules,
-    * since at least 2.6.0.
-    */
-
-   extern unsigned long totalram_pages;
-
-   unsigned int totalPhysicalPages = totalram_pages;
+   unsigned int totalPhysicalPages = compat_totalram_pages();

 #if LINUX_VERSION_CODE < KERNEL_VERSION(2, 6, 28)
   return MemDefaults_CalcMaxLockedPages(totalPhysicalPages);
@@ -1589,16 +1667,11 @@ HostIF_EstimateLockedPageLimit(const VMDriver* vm,                // IN
   unsigned int reservedPages = MEMDEFAULTS_MIN_HOST_PAGES;
   unsigned int hugePages = (vm == NULL) ? 0 :
      BYTES_2_PAGES(vm->memInfo.hugePageBytes);
-   unsigned int lockedPages = global_page_state(NR_PAGETABLE) +
-                              global_page_state(NR_SLAB_UNRECLAIMABLE) +
-                              global_page_state(NR_UNEVICTABLE) +
+   unsigned int lockedPages = global_zone_page_state(NR_PAGETABLE) +
+                              get_nr_slab_unreclaimable() +
+                              get_nr_unevictable() +
                              hugePages + reservedPages;
-   unsigned int anonPages =
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 8, 0)
-      global_page_state(NR_ANON_MAPPED);
-#else
-      global_page_state(NR_ANON_PAGES);
-#endif
+   unsigned int anonPages = get_nr_anon_mapped();
   unsigned int swapPages = BYTES_2_PAGES(linuxState.swapSize);

   if (anonPages > swapPages) {
@@ -1656,146 +1729,6 @@ HostIF_WaitForFreePages(unsigned int timeoutMs)  // IN:
 }


-/*
- *----------------------------------------------------------------------
- *
- * HostIFReadUptimeWork --
- *
- *      Reads the current uptime.  The uptime is based on getimeofday,
- *      which provides the needed high resolution.  However, we don't
- *      want uptime to be warped by e.g. calls to settimeofday.  So, we
- *      use a jiffies based monotonic clock to sanity check the uptime.
- *      If the uptime is more than one second from the monotonic time,
- *      we assume that the time of day has been set, and recalculate the
- *      uptime base to get uptime back on track with monotonic time.  On
- *      the other hand, we do expect jiffies based monotonic time and
- *      timeofday to have small drift (due to NTP rate correction, etc).
- *      We handle this by rebasing the jiffies based monotonic clock
- *      every second (see HostIFUptimeResyncMono).
- *      
- * Results:
- *      The uptime, in units of UPTIME_FREQ.  Also returns the jiffies
- *      value that was used in the monotonic time calculation.
- *
- * Side effects:
- *      May reset the uptime base in the case gettimeofday warp was 
- *      detected.
- *
- *----------------------------------------------------------------------
- */
-
-static uint64
-HostIFReadUptimeWork(unsigned long *j)  // OUT: current jiffies 
-{
-   struct timeval tv;
-   uint64 monotime, uptime, upBase, monoBase;
-   int64 diff;
-   uint32 version;
-   unsigned long jifs, jifBase;
-   unsigned int attempts = 0;
-
-   /* Assert that HostIF_InitUptime has been called. */
-   ASSERT(uptimeState.timer.function);
-
- retry:
-   do {
-      version  = VersionedAtomic_BeginTryRead(&uptimeState.version);
-      jifs     = jiffies;
-      jifBase  = uptimeState.jiffiesBase;
-      monoBase = uptimeState.monotimeBase;
-   } while (!VersionedAtomic_EndTryRead(&uptimeState.version, version));
-
-   do_gettimeofday(&tv);
-   upBase = Atomic_Read64(&uptimeState.uptimeBase);
-   
-   monotime = (uint64)(jifs - jifBase) * (UPTIME_FREQ / HZ);
-   monotime += monoBase;
-
-   uptime = tv.tv_usec * (UPTIME_FREQ / 1000000) + tv.tv_sec * UPTIME_FREQ;
-   uptime += upBase;
-   
-   /* 
-    * Use the jiffies based monotonic time to sanity check gettimeofday.
-    * If they differ by more than one second, assume the time of day has
-    * been warped, and use the jiffies time to undo (most of) the warp.
-    */
-
-   diff = uptime - monotime;
-   if (UNLIKELY(diff < -UPTIME_FREQ || diff > UPTIME_FREQ)) {
-      /* Compute a new uptimeBase to get uptime back on track. */
-      uint64 newUpBase = monotime - (uptime - upBase);
-
-      attempts++;
-      if (!Atomic_CMPXCHG64(&uptimeState.uptimeBase, &upBase, &newUpBase) && 
-          attempts < 5) {
-         /* Another thread updated uptimeBase.  Recalculate uptime. */
-         goto retry;
-      }
-      uptime = monotime;
-
-      Log("%s: detected settimeofday: fixed uptimeBase old %"FMT64"u "
-          "new %"FMT64"u attempts %u\n", __func__,
-          upBase, newUpBase, attempts);
-   }
-   *j = jifs;
-
-   return uptime;
-}
-
-
-/*
- *----------------------------------------------------------------------
- *
- * HostIFUptimeResyncMono --
- *
- *      Timer that fires ever second to resynchronize the jiffies based
- *      monotonic time with the uptime.
- *
- * Results:
- *      None
- *
- * Side effects:
- *      Resets the monotonic time bases so that jiffies based monotonic
- *      time does not drift from gettimeofday over the long term.
- *
- *----------------------------------------------------------------------
- */
-
-static void
-HostIFUptimeResyncMono(unsigned long data)  // IN: ignored
-{
-   unsigned long jifs;
-   uintptr_t flags;
-
-   /* 
-    * Read the uptime and the corresponding jiffies value.  This will
-    * also correct the uptime (which is based on time of day) if needed
-    * before we rebase monotonic time (which is based on jiffies).
-    */
-
-   uint64 uptime = HostIFReadUptimeWork(&jifs);
-
-   /* 
-    * Every second, recalculate monoBase and jiffiesBase to squash small
-    * drift between gettimeofday and jiffies.  Also, this prevents
-    * (jiffies - jiffiesBase) wrap on 32-bits.
-    */
-
-   SAVE_FLAGS(flags);
-   CLEAR_INTERRUPTS();
-   VersionedAtomic_BeginWrite(&uptimeState.version);
-
-   uptimeState.monotimeBase = uptime;
-   uptimeState.jiffiesBase  = jifs;
-
-   VersionedAtomic_EndWrite(&uptimeState.version);
-   RESTORE_FLAGS(flags);
-
-   /* Reschedule this timer to expire in one second. */
-   mod_timer(&uptimeState.timer, jifs + HZ);
-}
-
-
 /*
 *----------------------------------------------------------------------
 *
@@ -1807,8 +1740,7 @@ HostIFUptimeResyncMono(unsigned long data)  // IN: ignored
 *      None
 *
 * Side effects:
- *      Sets the initial values for the uptime state, and schedules
- *      the uptime timer.
+ *      Sets the initial value for the uptime base.
 *
 *----------------------------------------------------------------------
 */
@@ -1816,17 +1748,7 @@ HostIFUptimeResyncMono(unsigned long data)  // IN: ignored
 void
 HostIF_InitUptime(void)
 {
-   struct timeval tv;
-
-   uptimeState.jiffiesBase = jiffies;
-   do_gettimeofday(&tv);
-   Atomic_Write64(&uptimeState.uptimeBase, 
-                  -(tv.tv_usec * (UPTIME_FREQ / 1000000) + 
-                    tv.tv_sec * UPTIME_FREQ));
-
-   init_timer(&uptimeState.timer);
-   uptimeState.timer.function = HostIFUptimeResyncMono;
-   mod_timer(&uptimeState.timer, jiffies + HZ);
+   uptime_base = ktime_get_ns();
 }


@@ -1835,13 +1757,13 @@ HostIF_InitUptime(void)
 *
 * HostIF_CleanupUptime --
 *
- *      Cleanup uptime state, called at module unloading time.
+ *      No-op, left for backward compatibility.
 *
 * Results:
 *      None
 *
 * Side effects:
- *      Deschedule the uptime timer.
+ *      None
 *
 *----------------------------------------------------------------------
 */
@@ -1849,7 +1771,6 @@ HostIF_InitUptime(void)
 void
 HostIF_CleanupUptime(void)
 {
-   del_timer_sync(&uptimeState.timer);
 }


@@ -1873,9 +1794,10 @@ HostIF_CleanupUptime(void)
 uint64
 HostIF_ReadUptime(void)
 {
-   unsigned long jifs;
+   u64 tm;

-   return HostIFReadUptimeWork(&jifs);
+   tm = ktime_get_ns();
+   return (tm - uptime_base) / (NSEC_PER_SEC / UPTIME_FREQ);
 }


@@ -1996,15 +1918,15 @@ HostIF_MapCrossPage(VMDriver *vm, // IN
      return NULL;
   }
   vPgAddr = (VA) MapCrossPage(page);
-   HostIF_GlobalLock(16);
+   HostIF_VMLock(vm, 27);
   if (vm->vmhost->crosspagePagesCount >= MAX_INITBLOCK_CPUS) {
-      HostIF_GlobalUnlock(16);
+      HostIF_VMUnlock(vm, 27);
      UnmapCrossPage(page, (void*)vPgAddr);

      return NULL;
   }
   vm->vmhost->crosspagePages[vm->vmhost->crosspagePagesCount++] = page;
-   HostIF_GlobalUnlock(16);
+   HostIF_VMUnlock(vm, 27);

   ret = vPgAddr | (((VA)p) & (PAGE_SIZE - 1));

@@ -2222,6 +2144,27 @@ HostIF_VMLockIsHeld(VMDriver *vm) // IN

 #if defined(CONFIG_SMP) || defined(CONFIG_X86_UP_IOAPIC) || \
    defined(CONFIG_X86_UP_APIC) || defined(CONFIG_X86_LOCAL_APIC)
+
+#if COMPAT_LINUX_VERSION_CHECK_LT(5, 8, 0)
+static long compat_copy_from_kernel_nofault(void *dst, VA src, size_t size)
+{
+   mm_segment_t old_fs;
+   long ret;
+
+   old_fs = get_fs();
+   set_fs(KERNEL_DS);
+   ret = HostIF_CopyFromUser(dst, (const void *)src, size);
+   set_fs(old_fs);
+
+   return ret;
+}
+#else
+static long compat_copy_from_kernel_nofault(void *dst, VA src, size_t size)
+{
+   return copy_from_kernel_nofault(dst, (const void *)src, size);
+}
+#endif
+
 /*
 *----------------------------------------------------------------------
 *
@@ -2241,15 +2184,11 @@ HostIF_VMLockIsHeld(VMDriver *vm) // IN
 static Bool
 isVAReadable(VA r)  // IN:
 {
-   mm_segment_t old_fs;
   uint32 dummy;
-   int ret;
-   
-   old_fs = get_fs();
-   set_fs(get_ds());
+   long ret;
+
   r = APICR_TO_ADDR(r, APICR_VERSION);
-   ret = HostIF_CopyFromUser(&dummy, (void*)r, sizeof(dummy));
-   set_fs(old_fs);
+   ret = compat_copy_from_kernel_nofault(&dummy, r, sizeof(dummy));

   return ret == 0;
 }
@@ -2279,7 +2218,7 @@ SetVMAPICAddr(VMDriver *vm, // IN/OUT: driver state
   volatile void *hostapic;

   ASSERT_ON_COMPILE(APICR_SIZE <= PAGE_SIZE);
-   hostapic = (volatile void *) ioremap_nocache(ma, PAGE_SIZE);
+   hostapic = (volatile void *) ioremap(ma, PAGE_SIZE);
   if (hostapic) {
      if ((APIC_VERSIONREG(hostapic) & 0xF0) == 0x10) {
 	 vm->hostAPIC.base = (volatile uint32 (*)[4]) hostapic;
@@ -2435,44 +2374,41 @@ HostIF_SemaphoreWait(VMDriver *vm,   // IN:
                     uint64 *args)   // IN:
 {
   struct file *file;
-   mm_segment_t old_fs;
   int res;
   int waitFD = args[0];
   int timeoutms = args[2];
   uint64 value;
+   struct poll_wqueues table;
+   unsigned int mask;

   file = vmware_fget(waitFD);
   if (file == NULL) {
      return MX_WAITERROR;
   }

-   old_fs = get_fs();
-   set_fs(get_ds());
-
-   {
-      struct poll_wqueues table;
-      unsigned int mask;
-      
-      poll_initwait(&table);
-      current->state = TASK_INTERRUPTIBLE;
-      mask = file->f_op->poll(file, &table.pt);
-      if (!(mask & (POLLIN | POLLERR | POLLHUP))) {
-         vm->vmhost->vcpuSemaTask[vcpuid] = current;
-         schedule_timeout(timeoutms * HZ / 1000);  // convert to Hz
-         vm->vmhost->vcpuSemaTask[vcpuid] = NULL;
-      }
-      current->state = TASK_RUNNING;
-      poll_freewait(&table);
+   poll_initwait(&table);
+   current->state = TASK_INTERRUPTIBLE;
+   mask = compat_vfs_poll(file, &table.pt);
+   if (!(mask & (POLLIN | POLLERR | POLLHUP))) {
+      vm->vmhost->vcpuSemaTask[vcpuid] = current;
+      schedule_timeout(timeoutms * HZ / 1000);  // convert to Hz
+      vm->vmhost->vcpuSemaTask[vcpuid] = NULL;
   }
+   current->state = TASK_RUNNING;
+   poll_freewait(&table);

   /*
    * Userland only writes in multiples of sizeof(uint64). This will allow
    * the code to happily deal with a pipe or an eventfd. We only care about
    * reading no bytes (EAGAIN - non blocking fd) or sizeof(uint64).
+    *
+    * Upstream Linux changed the function parameter types/ordering in 4.14.0.
    */
-
-   res = file->f_op->read(file, (char *) &value, sizeof value, &file->f_pos);
-
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 14, 0)
+   res = kernel_read(file, file->f_pos, (char *)&value, sizeof value);
+#else
+   res = kernel_read(file, &value, sizeof value, &file->f_pos);
+#endif
   if (res == sizeof value) {
      res = MX_WAITNORMAL;
   } else {
@@ -2481,7 +2417,6 @@ HostIF_SemaphoreWait(VMDriver *vm,   // IN:
      }
   }

-   set_fs(old_fs);
   fput(file);

   /*
@@ -2564,8 +2499,8 @@ HostIF_SemaphoreForceWakeup(VMDriver *vm,       // IN:
 int
 HostIF_SemaphoreSignal(uint64 *args)  // IN:
 {
+   struct eventfd_ctx *eventfd;
   struct file *file;
-   mm_segment_t old_fs;
   int res;
   int signalFD = args[1];
   uint64 value = 1;  // make an eventfd happy should it be there
@@ -2575,22 +2510,34 @@ HostIF_SemaphoreSignal(uint64 *args)  // IN:
      return MX_WAITERROR;
   }

-   old_fs = get_fs();
-   set_fs(get_ds());
+   /*
+    * If it's eventfd, use specific eventfd interface as kernel writes
+    * to eventfd may not be allowed in kernel 5.10 and later.
+    */
+   eventfd = eventfd_ctx_fileget(file);
+   if (!IS_ERR(eventfd)) {
+      eventfd_signal(eventfd, 1);
+      fput(file);
+      return MX_WAITNORMAL;
+   }

   /*
    * Always write sizeof(uint64) bytes. This works fine for eventfd and
    * pipes. The data written is formatted to make an eventfd happy should
    * it be present.
+    *
+    * Upstream Linux changed the function parameter types/ordering in 4.14.0.
    */
-
-   res = file->f_op->write(file, (char *) &value, sizeof value, &file->f_pos);
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 14, 0)
+   res = kernel_write(file, (char *)&value, sizeof value, file->f_pos);
+#else
+   res = kernel_write(file, &value, sizeof value, &file->f_pos);
+#endif

   if (res == sizeof value) {
      res = MX_WAITNORMAL;
   }

-   set_fs(old_fs);
   fput(file);

   /*
@@ -2812,13 +2759,75 @@ HostIF_CallOnEachCPU(void (*func)(void*), // IN: function to call
 }


+/*
+ *-----------------------------------------------------------------------------
+ *
+ * HostIFCheckTrackedMPN --
+ *
+ *      Check if a given MPN is tracked for the specified VM.
+ *
+ * Result:
+ *      TRUE if the MPN is tracked in one of the trackers for the specified VM,
+ *      FALSE otherwise.
+ *
+ * Side effects:
+ *      None
+ *
+ *-----------------------------------------------------------------------------
+ */
+
+Bool
+HostIFCheckTrackedMPN(VMDriver *vm, // IN: The VM instance
+                      MPN mpn)      // IN: The MPN
+{
+   VMHost * const vmh = vm->vmhost;
+
+   if (vmh == NULL) {
+      return FALSE;
+   }
+
+   HostIF_VMLock(vm, 32); // Debug version of PhysTrack wants VM's lock.
+   if (vmh->lockedPages) {
+      if (PhysTrack_Test(vmh->lockedPages, mpn)) {
+         HostIF_VMUnlock(vm, 32);
+         return TRUE;
+      }
+   }
+
+   if (vmh->AWEPages) {
+      if (PhysTrack_Test(vmh->AWEPages, mpn)) {
+         HostIF_VMUnlock(vm, 32);
+         return TRUE;
+      }
+   }
+
+   if (vm->memtracker) {
+      if (MemTrack_LookupMPN(vm->memtracker, mpn) != NULL) {
+         HostIF_VMUnlock(vm, 32);
+         return TRUE;
+      }
+   }
+   HostIF_VMUnlock(vm, 32);
+
+   if (vmx86_debug) {
+      /*
+       * The monitor may have old KSeg mappings to pages which it no longer
+       * owns.  Minimize customer noise by only logging this for developers.
+       */
+      Log("%s: MPN %" FMT64 "x not owned by this VM\n", __FUNCTION__, mpn);
+   }
+   return FALSE;
+}
+
+
 /*
 *----------------------------------------------------------------------
 *
 * HostIF_ReadPage --
 *
- *      puts the content of a machine page into a kernel or user mode 
- *      buffer. 
+ *      Reads one page of data from a machine page and returns it in the
+ *      specified kernel or user buffer.  The machine page must be owned by
+ *      the specified VM.
 *
 * Results:
 *      0 on success
@@ -2831,7 +2840,8 @@ HostIF_CallOnEachCPU(void (*func)(void*), // IN: function to call
 */

 int
-HostIF_ReadPage(MPN mpn,             // MPN of the page
+HostIF_ReadPage(VMDriver *vm,        // IN: The VM instance
+                MPN mpn,             // MPN of the page
                VA64 addr,           // buffer for data
                Bool kernelBuffer)   // is the buffer in kernel space?
 {
@@ -2843,6 +2853,9 @@ HostIF_ReadPage(MPN mpn,             // MPN of the page
   if (mpn == INVALID_MPN) {
      return -EFAULT;
   }
+   if (HostIFCheckTrackedMPN(vm, mpn) == FALSE) {
+      return -EFAULT;
+   }

   page = pfn_to_page(mpn);
   ptr = kmap(page);
@@ -2866,8 +2879,8 @@ HostIF_ReadPage(MPN mpn,             // MPN of the page
 *
 * HostIF_WritePage --
 *
- *      Put the content of a kernel or user mode buffer into a machine 
- *      page.
+ *      Writes one page of data from a kernel or user buffer onto the specified
+ *      machine page.  The machine page must be owned by the specified VM.
 *
 * Results:
 *      0 on success
@@ -2880,9 +2893,9 @@ HostIF_ReadPage(MPN mpn,             // MPN of the page
 */

 int
-HostIF_WritePage(MPN mpn,              // MPN of the page
-                 VA64 addr,            // data to write to the page
-                 Bool kernelBuffer)    // is the buffer in kernel space?
+HostIFWritePageWork(MPN mpn,              // MPN of the page
+                    VA64 addr,            // data to write to the page
+                    Bool kernelBuffer)    // is the buffer in kernel space?
 {
   void const *buf = VA64ToPtr(addr);
   int ret = 0;
@@ -2909,6 +2922,45 @@ HostIF_WritePage(MPN mpn,              // MPN of the page
   return ret;
 }

+int
+HostIF_WritePage(VMDriver *vm,      // IN: The VM instance
+                 MPN mpn,              // MPN of the page
+                 VA64 addr,            // data to write to the page
+                 Bool kernelBuffer)    // is the buffer in kernel space?
+{
+   if (HostIFCheckTrackedMPN(vm, mpn) == FALSE) {
+      return -EFAULT;
+   }
+   return HostIFWritePageWork(mpn, addr, kernelBuffer);
+}
+
+
+/*
+ *-----------------------------------------------------------------------------
+ *
+ * HostIF_WriteMachinePage --
+ *
+ *      Puts the content of a machine page into a kernel or user mode
+ *      buffer.  This should only be used for host-global pages, not any
+ *      VM-owned pages.
+ *
+ * Results:
+ *      On success: 0
+ *      On failure: a negative error code
+ *
+ * Side effects:
+ *      None
+ *
+ *-----------------------------------------------------------------------------
+ */
+
+int
+HostIF_WriteMachinePage(MPN mpn,   // IN: MPN of the page
+                        VA64 addr) // IN: data to write to the page
+{
+   return HostIFWritePageWork(mpn, addr, TRUE);
+}
+

 /*
 *----------------------------------------------------------------------
@@ -3215,7 +3267,11 @@ HostIFStartTimer(Bool rateChanged,  //IN: Did rate change?
         return -1;
      }
   }
-   res = filp->f_op->read(filp, (void *) &buf, sizeof(buf), &pos);
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 14, 0)
+   res = kernel_read(filp, pos, (char *)&buf, sizeof(buf));
+#else
+   res = kernel_read(filp, &buf, sizeof(buf), &pos);
+#endif
   if (res <= 0) {
      if (res != -ERESTARTSYS) {
         Log("/dev/rtc read failed: %d\n", res);
@@ -3254,12 +3310,9 @@ HostIFFastClockThread(void *data)  // IN:
 {
   struct file *filp = (struct file *) data;
   int res;
-   mm_segment_t oldFS;
   unsigned int rate = 0;
   unsigned int prevRate = 0;

-   oldFS = get_fs();
-   set_fs(KERNEL_DS);
   allow_signal(SIGKILL);
   set_user_nice(current, linuxState.fastClockPriority);

@@ -3293,7 +3346,6 @@ HostIFFastClockThread(void *data)  // IN:

 out:
   LinuxDriverWakeUp(TRUE);
-   set_fs(oldFS);

   /*
    * Do not exit thread until we are told to do so.
@@ -3414,7 +3466,7 @@ HostIF_SetFastClockRate(unsigned int rate) // IN: Frequency in Hz.
      }
   } else {
      if (linuxState.fastClockThread) {
-         force_sig(SIGKILL, linuxState.fastClockThread);
+         send_sig(SIGKILL, linuxState.fastClockThread, 1);
         kthread_stop(linuxState.fastClockThread);
 	 close_rtc(linuxState.fastClockFile, current->files);

@@ -3462,7 +3514,7 @@ HostIF_MapUserMem(VA addr,                  // IN: User memory virtual address

   ASSERT(handle);

-   if (!access_ok(VERIFY_WRITE, p, size)) {
+   if (!write_access_ok(p, size)) {
      printk(KERN_ERR "%s: Couldn't verify write to uva 0x%p with size %"
             FMTSZ"u\n", __func__, p, size);

@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998-2015 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998-2017 VMware, Inc. All rights reserved.
 *
 * 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
@@ -205,11 +205,13 @@ skipTaskSwitch:;
         uint32 nPages = (uint32)crosspage->args[1];
         VA64   uAddr  = (VA64)VPN_2_VA(vpn);
         ASSERT(nPages <= MODULECALL_NUM_ARGS);
+         HostIF_VMLock(vm, 38);
         for (i = 0; i < nPages; i++) {
            MPN mpn;
            HostIF_LookupUserMPN(vm, uAddr + i * PAGE_SIZE, &mpn);
            crosspage->args[i] = mpn;
         }
+         HostIF_VMUnlock(vm, 38);
         break;
      }

@@ -117,7 +117,7 @@ prebuild:: ;
 postbuild:: ;

 $(DRIVER_KO): prebuild
-	$(MAKE) -C $(BUILD_DIR) SUBDIRS=$$PWD SRCROOT=$$PWD/$(SRCROOT) \
+	$(MAKE) -C $(BUILD_DIR) M=$$PWD SRCROOT=$$PWD/$(SRCROOT) \
 	  MODULEBUILDDIR=$(MODULEBUILDDIR) modules
 	$(MAKE) -C $$PWD SRCROOT=$$PWD/$(SRCROOT) \
 	  MODULEBUILDDIR=$(MODULEBUILDDIR) postbuild
@@ -35,6 +35,6 @@ $(DRIVER)-y := driver.o hub.o userif.o netif.o bridge.o procfs.o smac_compat.o \
 ####

 clean:
-	rm -rf $(DRIVER).o $(DRIVER).mod.o $(DRIVER).mod.c $(DRIVER).ko \
+	rm -rf $(DRIVER).o $(DRIVER).mod.o $(DRIVER).mod.c $(DRIVER).ko .cache.mk \
 	       .tmp_versions Module.symvers Modules.symvers Module.markers \
 	       modules.order $($(DRIVER)-y) .*.cmd .*.o.flags
@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998-2013 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998-2013, 2017 VMware, Inc. All rights reserved.
 *
 * 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
@@ -65,7 +65,7 @@
 #endif

 #if LOGLEVEL >= 4
-static struct timeval vnetTime;
+static u64 vnetTime;
 #endif

 typedef struct VNetBridge VNetBridge;
@@ -636,7 +636,7 @@ VNetBridgeReceiveFromVNet(VNetJack        *this, // IN: jack
 	 unsigned long flags;
 	 int i;

-	 atomic_inc(&clone->users);
+	 clone = skb_get(clone);

 	 clone->dev = dev;
 	 clone->protocol = eth_type_trans(clone, dev);
@@ -696,7 +696,7 @@ VNetBridgeReceiveFromVNet(VNetJack        *this, // IN: jack

 	 netif_rx_ni(clone);
 #	 if LOGLEVEL >= 4
-	 do_gettimeofday(&vnetTime);
+	 vnetTime = ktime_get_ns();
 #	 endif
      }
   }
@@ -1104,7 +1104,7 @@ VNetBridgeNotifyLogBridgeUpError(int errno,        // IN: the error number
      case -ENODEV:
         LOG(0, (KERN_WARNING "bridge-%s: interface %s not found or not "
                 "up\n", bridgeName, devName));
-	 break;
+         break;
      case -EINVAL:
         LOG(0, (KERN_WARNING "bridge-%s: interface %s is not a valid "
                 "Ethernet interface\n", bridgeName, devName));
@@ -1143,11 +1143,16 @@ VNetBridgeNotifyLogBridgeUpError(int errno,        // IN: the error number
 static int
 VNetBridgeNotify(struct notifier_block *this, // IN: callback data (bridge)
                 u_long msg,                  // IN: type of event
-                 void *data)                  // IN: device pertaining to event
+                 void *data)                  // IN: net_device or notifier info
 {
   VNetBridge *bridge = list_entry(this, VNetBridge, notifier);
-   struct net_device *dev = (struct net_device *) data;
+   struct net_device *dev;

+#if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 11, 0)
+   dev = netdev_notifier_info_to_dev(data);
+#else
+   dev = (struct net_device *)data;
+#endif
   switch (msg) {
   case NETDEV_UNREGISTER:
      LOG(2, (KERN_DEBUG "bridge-%s: interface %s is unregistering\n",
@@ -1179,7 +1184,7 @@ VNetBridgeNotify(struct notifier_block *this, // IN: callback data (bridge)

         LOG(1, (KERN_DEBUG "bridge-%s: enabling the bridge on dev up\n",
                 bridge->name));
-	 errno = VNetBridgeUp(bridge, FALSE);
+         errno = VNetBridgeUp(bridge, FALSE);
         if (errno != 0) {
            VNetBridgeNotifyLogBridgeUpError(errno, bridge->name, dev->name);
         }
@@ -1656,12 +1661,11 @@ VNetBridgeReceiveFromDev(struct sk_buff *skb,         // IN: packet to receive

 #  if LOGLEVEL >= 4
   {
-      struct timeval now;
-      do_gettimeofday(&now);
+      u64 now;
+
+      now = ktime_get_ns();
      LOG(3, (KERN_DEBUG "bridge-%s: time %d\n",
-	      bridge->name,
-	      (int)((now.tv_sec * 1000000 + now.tv_usec)
-                    - (vnetTime.tv_sec * 1000000 + vnetTime.tv_usec))));
+	      bridge->name, (int)((now - vnetTime) / NSEC_PER_USEC)));
   }
 #  endif

@@ -0,0 +1,34 @@
+#ifndef __COMPAT_MMAP_LOCK_H__
+#define __COMPAT_MMAP_LOCK_H__
+
+#include <linux/mm.h>
+
+/*
+ * In 5.8-rc1, mmap locking was reworked to use wrappers around mmap_sem
+ * (which was also renamed to mmap_lock). All code is now supposed to use
+ * these wrappers as the internal implementation of the lock may change in
+ * the future.
+ *
+ * Check also _LINUX_MMAP_LOCK_H to handle possible backports to distribution
+ * pre-5.8 kernel. This macro is defined in <linux/mmap_lock.h> which is also
+ * included in <linux/mm.h> since the commit introducing the wrappers so that
+ * we should have it defined in any kernel providing the new API.
+ */
+
+#if (LINUX_VERSION_CODE >= KERNEL_VERSION(5, 8, 0)) || defined(_LINUX_MMAP_LOCK_H)
+#include <linux/mmap_lock.h>
+#else
+
+static inline void mmap_read_lock(struct mm_struct *mm)
+{
+	down_read(&mm->mmap_sem);
+}
+
+static inline void mmap_read_unlock(struct mm_struct *mm)
+{
+	up_read(&mm->mmap_sem);
+}
+
+#endif /* 5.8.0 */
+
+#endif /* __COMPAT_MMAP_LOCK_H__ */
@@ -337,7 +337,8 @@ typedef netdev_features_t compat_netdev_features_t;
 typedef u32 compat_netdev_features_t;
 #endif

-#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 7, 0)
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 7, 0) || \
+    (defined(RHEL_RELEASE_CODE) && RHEL_RELEASE_CODE >= 0x0704)
 #define compat_netif_trans_update(d) netif_trans_update(d)
 #else
 #define compat_netif_trans_update(d) do { (d)->trans_start = jiffies; } while (0)
@@ -161,6 +161,14 @@ VNetProcOpen(struct inode *inode,   // IN:
 }

 /* Our procfs callbacks.  We only need to specialize open. */
+#ifdef DEFINE_PROC_SHOW_ATTRIBUTE
+static const struct proc_ops fops = {
+   .proc_open    = VNetProcOpen,
+   .proc_read    = seq_read,
+   .proc_lseek   = seq_lseek,
+   .proc_release = single_release,
+};
+#else
 static struct file_operations fops = {
   .open    = VNetProcOpen,
   .read    = seq_read,
@@ -168,6 +176,7 @@ static struct file_operations fops = {
   .release = single_release,
 };
 #endif
+#endif


 /*
@@ -36,6 +36,9 @@
 #include <linux/slab.h>
 #include <linux/version.h>
 #include <linux/wait.h>
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 11, 0)
+#include <linux/taskstats_kern.h>  // For <linux/sched/signal.h> without version dependency
+#endif

 #include <net/checksum.h>
 #include <net/sock.h>
@@ -44,6 +47,7 @@

 #include "vnetInt.h"
 #include "compat_skbuff.h"
+#include "compat_mmap_lock.h"
 #include "vmnetInt.h"
 #include "vm_atomic.h"
 #include "vm_assert.h"
@@ -81,11 +85,55 @@ static int  VNetUserIfSetUplinkState(VNetPort *port, uint8 linkUp);
 extern unsigned int  vnet_max_qlen;

 #if COMPAT_LINUX_VERSION_CHECK_LT(3, 2, 0)
-#   define compat_kmap(page) kmap(page)
-#   define compat_kunmap(page) kunmap(page)
+#   define compat_kmap_frag(frag) kmap((frag)->page)
+#   define compat_kunmap_frag(page) kunmap((frag)->page)
 #else
-#   define compat_kmap(page) kmap((page).p)
-#   define compat_kunmap(page) kunmap((page).p)
+#   define compat_kmap_frag(frag) kmap(skb_frag_page(frag))
+#   define compat_kunmap_frag(frag) kunmap(skb_frag_page(frag))
+#endif
+
+static unsigned int compat_skb_frag_size(const skb_frag_t *frag)
+{
+#if COMPAT_LINUX_VERSION_CHECK_LT(3, 2, 0)
+	return frag->size;
+#else
+	return skb_frag_size(frag);
+#endif
+}
+
+static unsigned int compat_skb_frag_off(const skb_frag_t *frag)
+{
+#if COMPAT_LINUX_VERSION_CHECK_LT(5, 4, 0) && \
+	!(defined(CONFIG_SUSE_VERSION) && CONFIG_SUSE_VERSION == 15 && \
+	  defined(CONFIG_SUSE_PATCHLEVEL) && CONFIG_SUSE_PATCHLEVEL >= 2)
+	return frag->page_offset;
+#else
+	return skb_frag_off(frag);
+#endif
+}
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(5, 0, 0) && defined(VERIFY_WRITE)
+	#define write_access_ok(addr, size) access_ok(VERIFY_WRITE, addr, size)
+#else
+	#define write_access_ok(addr, size) access_ok(addr, size)
+#endif
+
+#if COMPAT_LINUX_VERSION_CHECK_LT(5, 10, 0)
+static inline
+__wsum compat_csum_and_copy_to_user(const void *src, void __user *dst, int len)
+{
+	int err;
+	__wsum ret;
+
+	ret = csum_and_copy_to_user(src, dst, len, 0, &err);
+	return err ? 0 : ret;
+}
+#else
+static inline
+__wsum compat_csum_and_copy_to_user(const void *src, void __user *dst, int len)
+{
+	return csum_and_copy_to_user(src, dst, len);
+}
 #endif

 /*
@@ -112,14 +160,16 @@ UserifLockPage(VA addr) // IN
   struct page *page = NULL;
   int retval;

-   down_read(&current->mm->mmap_sem);
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 6, 0)
+   mmap_read_lock(current->mm);
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 9, 0)
+   retval = get_user_pages(addr, 1, FOLL_WRITE, &page, NULL);
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(4, 6, 0)
   retval = get_user_pages(addr, 1, 1, 0, &page, NULL);
 #else
   retval = get_user_pages(current, current->mm, addr,
                           1, 1, 0, &page, NULL);
 #endif
-   up_read(&current->mm->mmap_sem);
+   mmap_read_unlock(current->mm);

   if (retval != 1) {
      return NULL;
@@ -152,7 +202,7 @@ VNetUserIfMapPtr(VA uAddr,        // IN: pointer to user memory
                 struct page **p, // OUT: locked page
                 void **ptr)      // OUT: kernel mapped pointer
 {
-   if (!access_ok(VERIFY_WRITE, (void *)uAddr, size) ||
+   if (!write_access_ok((void *)uAddr, size) ||
       (((uAddr + size - 1) & ~(PAGE_SIZE - 1)) !=
        (uAddr & ~(PAGE_SIZE - 1)))) {
      return -EINVAL;
@@ -567,7 +617,6 @@ VNetCsumCopyDatagram(const struct sk_buff *skb,	// IN: skb to copy
 		     char *buf)			// OUT: where to copy data
 {
   unsigned int csum;
-   int err = 0;
   int len = skb_headlen(skb) - offset;
   char *curr = buf;
   const skb_frag_t *frag;
@@ -580,29 +629,27 @@ VNetCsumCopyDatagram(const struct sk_buff *skb,	// IN: skb to copy
      return -EINVAL;
   }

-   csum = csum_and_copy_to_user(skb->data + offset, curr, len, 0, &err);
-   if (err) {
-      return err;
-   }
+   csum = compat_csum_and_copy_to_user(skb->data + offset, curr, len);
+   if (!csum)
+	   return -EFAULT;
   curr += len;

   for (frag = skb_shinfo(skb)->frags;
 	frag != skb_shinfo(skb)->frags + skb_shinfo(skb)->nr_frags;
 	frag++) {
-      if (frag->size > 0) {
+      if (compat_skb_frag_size(frag) > 0) {
 	 unsigned int tmpCsum;
 	 const void *vaddr;

-	 vaddr = compat_kmap(frag->page);
-	 tmpCsum = csum_and_copy_to_user(vaddr + frag->page_offset,
-					 curr, frag->size, 0, &err);
-	 compat_kunmap(frag->page);
+	 vaddr = compat_kmap_frag(frag);
+	 tmpCsum = compat_csum_and_copy_to_user(vaddr + compat_skb_frag_off(frag),
+						curr, compat_skb_frag_size(frag));
+	 compat_kunmap_frag(frag);

-	 if (err) {
-	    return err;
-	 }
+	 if (!tmpCsum)
+		 return -EFAULT;
 	 csum = csum_block_add(csum, tmpCsum, curr - buf);
-	 curr += frag->size;
+	 curr += compat_skb_frag_size(frag);
      }
   }

@@ -35,6 +35,8 @@
 #endif
 #endif

+#include <linux/pci_ids.h>
+
 /* LSILogic 53C1030 Parallel SCSI controller
 * LSILogic SAS1068 SAS controller
 */
@@ -53,7 +55,10 @@
 *    VMware HD Audio codec
 *    VMware HD Audio controller
 */
+#ifndef PCI_VENDOR_ID_VMWARE
 #define PCI_VENDOR_ID_VMWARE                    0x15AD
+#endif
+
 #define PCI_DEVICE_ID_VMWARE_SVGA2              0x0405
 #define PCI_DEVICE_ID_VMWARE_SVGA               0x0710
 #define PCI_DEVICE_ID_VMWARE_VGA                0x0711
@@ -70,7 +75,11 @@
 #define PCI_DEVICE_ID_VMWARE_1394               0x0780
 #define PCI_DEVICE_ID_VMWARE_BRIDGE             0x0790
 #define PCI_DEVICE_ID_VMWARE_ROOTPORT           0x07A0
+
+#ifndef PCI_DEVICE_ID_VMWARE_VMXNET3
 #define PCI_DEVICE_ID_VMWARE_VMXNET3            0x07B0
+#endif
+
 #define PCI_DEVICE_ID_VMWARE_PVSCSI             0x07C0
 #define PCI_DEVICE_ID_VMWARE_82574              0x07D0
 #define PCI_DEVICE_ID_VMWARE_AHCI               0x07E0
@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998, 2017 VMware, Inc. All rights reserved.
 *
 * 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
@@ -77,7 +77,7 @@


 extern struct proto vmnet_proto;
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 2, 0)
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 2, 0) || defined(sk_net_refcnt)
 #   define compat_sk_alloc(_bri, _pri) sk_alloc(&init_net, \
                                                PF_NETLINK, _pri, &vmnet_proto, 1)
 #elif defined(VMW_NETDEV_HAS_NET)
@@ -1,43 +0,0 @@
-/*********************************************************
- * Copyright (C) 2002-2015 VMware, Inc. All rights reserved.
- *
- * 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 version 2 and no 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 St, Fifth Floor, Boston, MA  02110-1301 USA
- *
- *********************************************************/
-
-/*
- * vmware_pack_begin.h --
- *
- *    Begin of structure packing. See vmware_pack_init.h for details.
- *
- *    Note that we do not use the following construct in this include file,
- *    because we want to emit the code every time the file is included --hpreg
- *
- *    #ifndef foo
- *    #   define foo
- *    ...
- *    #endif
- *
- */
-
-
-#include "vmware_pack_init.h"
-
-
-#ifdef _MSC_VER
-#   pragma pack(push, 1)
-#elif __GNUC__
-#else
-#   error Compiler packing...
-#endif
@@ -1,44 +0,0 @@
-/*********************************************************
- * Copyright (C) 2002-2015 VMware, Inc. All rights reserved.
- *
- * 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 version 2 and no 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 St, Fifth Floor, Boston, MA  02110-1301 USA
- *
- *********************************************************/
-
-/*
- * vmware_pack_end.h --
- *
- *    End of structure packing. See vmware_pack_init.h for details.
- *
- *    Note that we do not use the following construct in this include file,
- *    because we want to emit the code every time the file is included --hpreg
- *
- *    #ifndef foo
- *    #   define foo
- *    ...
- *    #endif
- *
- */
-
-
-#include "vmware_pack_init.h"
-
-
-#ifdef _MSC_VER
-#   pragma pack(pop)
-#elif __GNUC__
-__attribute__((__packed__))
-#else
-#   error Compiler packing...
-#endif
@@ -1,65 +0,0 @@
-/*********************************************************
- * Copyright (C) 2002-2015 VMware, Inc. All rights reserved.
- *
- * 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 version 2 and no 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 St, Fifth Floor, Boston, MA  02110-1301 USA
- *
- *********************************************************/
-
-#ifndef __VMWARE_PACK_INIT_H__
-#   define __VMWARE_PACK_INIT_H__
-
-
-/*
- * vmware_pack_init.h --
- *
- *    Platform-independent code to make the compiler pack (i.e. have them
- *    occupy the smallest possible space) structure definitions. The following
- *    constructs are known to work --hpreg
- *
- *    #include "vmware_pack_begin.h"
- *    struct foo {
- *       ...
- *    }
- *    #include "vmware_pack_end.h"
- *    ;
- *
- *    typedef
- *    #include "vmware_pack_begin.h"
- *    struct foo {
- *       ...
- *    }
- *    #include "vmware_pack_end.h"
- *    foo;
- */
-
-
-#ifdef _MSC_VER
-/*
- * MSVC 6.0 emits warning 4103 when the pack push and pop pragma pairing is
- * not balanced within 1 included file. That is annoying because our scheme
- * is based on the pairing being balanced between 2 included files.
- *
- * So we disable this warning, but this is safe because the compiler will also
- * emit warning 4161 when there is more pops than pushes within 1 main
- * file --hpreg
- */
-
-#   pragma warning(disable:4103)
-#elif __GNUC__
-#else
-#   error Compiler packing...
-#endif
-
-
-#endif /* __VMWARE_PACK_INIT_H__ */