vmmon: get rid of do_gettimeofday() in linux/driver.c

The do_gettimeofday() helper was removed by commit e4b92b108c6c
("timekeeping: remove obsolete time accessors") in v5.0-rc1 and commit
c766d1472c70 ("y2038: hide timeval/timespec/itimerval/itimerspec types")
in v5.6-rc3 hides struct timeval from kernel code.

Replace the use of do_gettimeofday() and struct timeval in linux/driver.c.
While ktime_get_real_ns() would match do_gettimeofday() better, monotonic
timer seems more appropriate for poll timeouts so use ktime_get_ns()
instead.

INSTALL

@@ -150,7 +150,7 @@ test build).
 When using a downloaded tarball, simply run
 
   tar -cf vmmon.tar vmmon-only
-  tar -cf vmnet.tar vmmon-only
+  tar -cf vmnet.tar vmnet-only
 
 to create the tarballs. In this case, there is no protection against
 unwanted local changes.

Makefile

@@ -43,6 +43,7 @@ install: retiredcheck $(MODFILES)
 	if test -z "$(DESTDIR)"; then $(DEPMOD) -a $(VM_UNAME); fi
 
 clean: $(SUBDIRS)
+	rm -f *.o
 
 tarballs: $(TARBALLS)
 

RETIRED

@@ -1,7 +0,0 @@
-*-----------------------------------------------------------------------------
-* Branch workstation-12.5.8 has been retired on 2018-01-18 and has not been
-* receiving any updates since. Please switch to latest 12.5.x version
-* available.
-*
-* Latest kernel version this branch is intended to work with is 4.14.
-*-----------------------------------------------------------------------------

vmmon-only/Makefile

@@ -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

vmmon-only/Makefile.kernel

@@ -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)))

vmmon-only/common/cpuid.c

@@ -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;
+}

vmmon-only/common/cpuid.h

@@ -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)

vmmon-only/common/task.c

@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998-2017 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
@@ -79,6 +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)
@@ -1594,12 +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)
    /*
@@ -1764,6 +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);
@@ -1973,6 +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);
@@ -1980,6 +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.
@@ -2113,6 +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

vmmon-only/include/compat_poll.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__ */

vmmon-only/include/hashFunc.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 */
   }

vmmon-only/include/iocontrols.h

@@ -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))
 

vmmon-only/include/modulecall.h

@@ -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)

vmmon-only/include/x86cpuid.h

@@ -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 \

vmmon-only/include/x86cpuid_asm.h

@@ -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
 

vmmon-only/include/x86msr.h

@@ -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
@@ -80,6 +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
@@ -107,6 +108,7 @@ MSRQuery;
 #define MSR_TSC_AUX           0xc0000103
 #define MSR_BD_TSC_RATIO      0xc0000104
 
+
 #ifndef MSR_MISC_FEATURES_ENABLES
 #define MSR_MISC_FEATURES_ENABLES            0x140
 #endif
@@ -303,7 +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
@@ -345,6 +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.

vmmon-only/linux/driver.c

@@ -81,6 +81,9 @@ static Bool LinuxDriverCheckPadding(void);
 
 struct VMXLinuxState linuxState;
 
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 17, 0)
+typedef int vm_fault_t;
+#endif
 
 /*
  *----------------------------------------------------------------------
@@ -106,9 +109,9 @@ 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 LINUX_VERSION_CODE >= KERNEL_VERSION(4, 11, 0)
-static int LinuxDriverFault(struct vm_fault *fault);
+static vm_fault_t LinuxDriverFault(struct vm_fault *fault);
 #elif defined(VMW_NOPAGE_2624)
-static int LinuxDriverFault(struct vm_area_struct *vma, struct vm_fault *fault);
+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,
@@ -738,14 +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;
@@ -812,12 +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) {
@@ -881,11 +880,11 @@ LinuxDriverPollTimeout(compat_timer_arg_t unused)  // IN:
  */
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 11, 0)
-static int
+static vm_fault_t
 LinuxDriverFault(struct vm_fault *fault)     //IN/OUT
 #elif defined(VMW_NOPAGE_2624)
-static int LinuxDriverFault(struct vm_area_struct *vma, //IN
-                            struct vm_fault *fault)     //IN/OUT
+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
@@ -1284,7 +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;

vmmon-only/linux/hostif.c

@@ -99,6 +99,7 @@
 #include "pgtbl.h"
 #include "vmmonInt.h"
 #include "versioned_atomic.h"
+#include "compat_poll.h"
 
 #if LINUX_VERSION_CODE < KERNEL_VERSION(4, 14, 0)
 #   define global_zone_page_state global_page_state
@@ -174,13 +175,7 @@ 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
@@ -239,6 +234,21 @@ 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
+
 /*
  *-----------------------------------------------------------------------------
  *
@@ -1601,14 +1611,7 @@ 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);
@@ -1688,146 +1691,6 @@ 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(compat_timer_arg_t unused)  // 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);
-}
-
-
 /*
  *----------------------------------------------------------------------
  *
@@ -1839,8 +1702,7 @@ HostIFUptimeResyncMono(compat_timer_arg_t unused)  // 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.
  *
  *----------------------------------------------------------------------
  */
@@ -1848,16 +1710,7 @@ HostIFUptimeResyncMono(compat_timer_arg_t unused)  // 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));
-
-   compat_timer_setup(&uptimeState.timer, HostIFUptimeResyncMono, 0);
-   mod_timer(&uptimeState.timer, jiffies + HZ);
+   uptime_base = ktime_get_ns();
 }
 
 
@@ -1866,13 +1719,13 @@ 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
  *
  *----------------------------------------------------------------------
  */
@@ -1880,7 +1733,6 @@ HostIF_InitUptime(void)
 void
 HostIF_CleanupUptime(void)
 {
-   del_timer_sync(&uptimeState.timer);
 }
 
 
@@ -1904,9 +1756,10 @@ 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);
 }
 
 
@@ -2277,7 +2130,7 @@ isVAReadable(VA r)  // IN:
    int ret;
    
    old_fs = get_fs();
-   set_fs(get_ds());
+   set_fs(KERNEL_DS);
    r = APICR_TO_ADDR(r, APICR_VERSION);
    ret = HostIF_CopyFromUser(&dummy, (void*)r, sizeof(dummy));
    set_fs(old_fs);
@@ -2310,7 +2163,7 @@ 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;
@@ -2478,7 +2331,7 @@ HostIF_SemaphoreWait(VMDriver *vm,   // IN:
    }
 
    old_fs = get_fs();
-   set_fs(get_ds());
+   set_fs(KERNEL_DS);
 
    {
       struct poll_wqueues table;
@@ -2486,7 +2339,7 @@ HostIF_SemaphoreWait(VMDriver *vm,   // IN:
       
       poll_initwait(&table);
       current->state = TASK_INTERRUPTIBLE;
-      mask = file->f_op->poll(file, &table.pt);
+      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
@@ -2607,7 +2460,7 @@ HostIF_SemaphoreSignal(uint64 *args)  // IN:
    }
 
    old_fs = get_fs();
-   set_fs(get_ds());
+   set_fs(KERNEL_DS);
 
    /*
     * Always write sizeof(uint64) bytes. This works fine for eventfd and
@@ -3550,7 +3403,7 @@ 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);
 
@@ -3598,7 +3451,7 @@ 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);
 

vmnet-only/Makefile

@@ -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

vmnet-only/Makefile.kernel

@@ -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

vmnet-only/bridge.c

@@ -65,7 +65,7 @@
 #endif
 
 #if LOGLEVEL >= 4
-static struct timeval vnetTime;
+static u64 vnetTime;
 #endif
 
 typedef struct VNetBridge VNetBridge;
@@ -696,7 +696,7 @@ VNetBridgeReceiveFromVNet(VNetJack        *this, // IN: jack
 
 	 netif_rx_ni(clone);
 #	 if LOGLEVEL >= 4
-	 do_gettimeofday(&vnetTime);
+	 vnetTime = ktime_get_ns();
 #	 endif
       }
    }
@@ -1661,12 +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
 

vmnet-only/procfs.c

@@ -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
 
 
 /*

vmnet-only/userif.c

@@ -84,11 +84,37 @@ 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
 
 /*
@@ -157,7 +183,7 @@ 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;
@@ -594,20 +620,21 @@ VNetCsumCopyDatagram(const struct sk_buff *skb,	// IN: skb to copy
    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 = csum_and_copy_to_user(vaddr + compat_skb_frag_off(frag),
+					 curr, compat_skb_frag_size(frag), 0,
+					 &err);
+	 compat_kunmap_frag(frag);
 
 	 if (err) {
 	    return err;
 	 }
 	 csum = csum_block_add(csum, tmpCsum, curr - buf);
-	 curr += frag->size;
+	 curr += compat_skb_frag_size(frag);
       }
    }
 

vmnet-only/vmware_pack_begin.h

@@ -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

vmnet-only/vmware_pack_end.h

@@ -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

vmnet-only/vmware_pack_init.h

@@ -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__ */