import Player 14.1.3 module sources

INSTALL

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

vmmon-only/Makefile

@@ -107,7 +107,7 @@ prebuild:: ;
 postbuild:: ;
 
 $(DRIVER_KO): prebuild
-	$(MAKE) -C $(BUILD_DIR) M=$$PWD SRCROOT=$$PWD/$(SRCROOT) \
+	$(MAKE) -C $(BUILD_DIR) SUBDIRS=$$PWD SRCROOT=$$PWD/$(SRCROOT) \
 	  MODULEBUILDDIR=$(MODULEBUILDDIR) modules
 	$(MAKE) -C $$PWD SRCROOT=$$PWD/$(SRCROOT) \
 	  MODULEBUILDDIR=$(MODULEBUILDDIR) postbuild

vmmon-only/Makefile.kernel

@@ -31,7 +31,7 @@ $(DRIVER)-y := $(subst $(SRCROOT)/, , $(patsubst %.c, %.o, \
 		$(SRCROOT)/bootstrap/*.c)))
 
 clean:
-	rm -rf $(wildcard $(DRIVER).mod.c $(DRIVER).ko .tmp_versions .cache.mk \
+	rm -rf $(wildcard $(DRIVER).mod.c $(DRIVER).ko .tmp_versions \
 		Module.symvers Modules.symvers Module.markers modules.order \
 		$(foreach dir,linux/ common/ vmcore/ bootstrap/ \
 		./,$(addprefix $(dir),.*.cmd .*.o.flags *.o)))

vmmon-only/common/task.c

@@ -2203,23 +2203,12 @@ 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)
    /*

vmmon-only/include/compat_poll.h

@@ -1,30 +0,0 @@
-#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

@@ -251,7 +251,6 @@ 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/vm_assert.h

@@ -67,7 +67,6 @@ 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) \
@@ -108,7 +107,7 @@ NORETURN void Panic_NoSave(const char *fmt, ...) PRINTF_DECL(1, 2);
 } while(0)
 
 #else
-#define Panic panic
+NORETURN void Panic(const char *fmt, ...) PRINTF_DECL(1, 2);
 #endif
 
 void LogThrottled(uint32 *count, const char *fmt, ...) PRINTF_DECL(2, 3);

vmmon-only/include/x86_basic_defs.h

@@ -35,8 +35,6 @@
 #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
@@ -77,9 +75,7 @@
 #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

vmmon-only/include/x86msr.h

@@ -24,7 +24,6 @@
 
 #ifndef _X86MSR_H_
 #define _X86MSR_H_
-#include <asm/msr-index.h>
 #define INCLUDE_ALLOW_USERLEVEL
 #define INCLUDE_ALLOW_VMX
 
@@ -127,9 +126,7 @@ MSRQuery;
 #define MSR_SPEC_CTRL_STIBP                       (1UL << 1)
 #define MSR_PRED_CMD_IBPB                         (1UL << 0)
 
-#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
@@ -453,9 +450,7 @@ 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
@@ -633,11 +628,7 @@ 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
 
 
 

vmmon-only/linux/driver.c

@@ -73,9 +73,6 @@ static Bool LinuxDriverCheckPadding(void);
 
 struct VMXLinuxState linuxState;
 
-#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 17, 0)
-typedef int vm_fault_t;
-#endif
 
 /*
  *----------------------------------------------------------------------
@@ -100,9 +97,9 @@ long LinuxDriver_Ioctl(struct file *filp, u_int iocmd,
 
 static int LinuxDriver_Close(struct inode *inode, struct file *filp);
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 11, 0)
-static vm_fault_t LinuxDriverFault(struct vm_fault *fault);
+static int LinuxDriverFault(struct vm_fault *fault);
 #else
-static vm_fault_t LinuxDriverFault(struct vm_area_struct *vma, struct vm_fault *fault);
+static int LinuxDriverFault(struct vm_area_struct *vma, struct vm_fault *fault);
 #endif
 static int LinuxDriverMmap(struct file *filp, struct vm_area_struct *vma);
 
@@ -597,7 +594,7 @@ LinuxDriver_Close(struct inode *inode, // IN
  *-----------------------------------------------------------------------------
  */
 
-static vm_fault_t
+static int
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 11, 0)
 LinuxDriverFault(struct vm_fault *fault)     //IN/OUT
 #else
@@ -984,7 +981,7 @@ LinuxDriverReadTSC(void *data,   // OUT: TSC values
  *-----------------------------------------------------------------------------
  */
 
-__always_inline static Bool
+__attribute__((always_inline)) static Bool
 LinuxDriverSyncReadTSCs(uint64 *delta) // OUT: TSC max - TSC min
 {
    TSCDelta tscDelta;

vmmon-only/linux/hostif.c

@@ -47,16 +47,13 @@
 #include <asm/asm.h>
 #include <asm/io.h>
 #include <asm/page.h>
-#include <asm/tlbflush.h>
 #include <asm/uaccess.h>
-#include <asm/irq_vectors.h>
 #include <linux/capability.h>
 #include <linux/kthread.h>
 #include <linux/wait.h>
 #include <linux/hrtimer.h>
 #include <linux/signal.h>
 #include <linux/taskstats_kern.h> // For linux/sched/signal.h without version check
-#include <linux/eventfd.h>
 
 #include "vmware.h"
 #include "x86apic.h"
@@ -77,7 +74,6 @@
 
 #include "pgtbl.h"
 #include "versioned_atomic.h"
-#include "compat_poll.h"
 
 #if !defined(CONFIG_HIGH_RES_TIMERS)
 #error CONFIG_HIGH_RES_TIMERS required for acceptable performance
@@ -119,7 +115,13 @@
  */
 #define LOCKED_PAGE_SLACK 10000
 
-u64 uptime_base;
+static struct {
+   Atomic_uint64     uptimeBase;
+   VersionedAtomic   version;
+   uint64            monotimeBase;
+   unsigned long     jiffiesBase;
+   struct timer_list timer;
+} uptimeState;
 
 /*
  * First Page Locking strategy
@@ -178,21 +180,6 @@ u64 uptime_base;
 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
-
 /*
  *-----------------------------------------------------------------------------
  *
@@ -606,15 +593,6 @@ 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
 
 /*
  *----------------------------------------------------------------------
@@ -635,29 +613,7 @@ static int crosspage_set_exec(pte_t *pte, unsigned long addr, void *data)
 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
 }
 
 
@@ -1538,7 +1494,14 @@ unsigned int
 HostIF_EstimateLockedPageLimit(const VMDriver* vm,                // IN
                                unsigned int currentlyLockedPages) // IN
 {
-   unsigned int totalPhysicalPages = compat_totalram_pages();
+   /*
+    * This variable is available and exported to modules,
+    * since at least 2.6.0.
+    */
+
+   extern unsigned long totalram_pages;
+
+   unsigned int totalPhysicalPages = totalram_pages;
 
    /*
     * Use the memory information linux exports as of late for a more
@@ -1563,11 +1526,8 @@ HostIF_EstimateLockedPageLimit(const VMDriver* vm,                // IN
 #else
    lockedPages += global_page_state(NR_PAGETABLE);
 #endif
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 9, 0)
-   /* NR_SLAB_* converted to byte counters in 5.9 */
-   lockedPages += global_node_page_state_pages(NR_SLAB_UNRECLAIMABLE_B);
-#elif LINUX_VERSION_CODE >= KERNEL_VERSION(4, 13, 0)
    /* NR_SLAB_* moved from zone to node in 4.13. */
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 13, 0)
    lockedPages += global_node_page_state(NR_SLAB_UNRECLAIMABLE);
 #else
    lockedPages += global_page_state(NR_SLAB_UNRECLAIMABLE);
@@ -1639,6 +1599,143 @@ 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;
+
+ 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(struct timer_list *timer)  // 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);
+}
+
+
 /*
  *----------------------------------------------------------------------
  *
@@ -1650,7 +1747,8 @@ HostIF_WaitForFreePages(unsigned int timeoutMs)  // IN:
  *      None
  *
  * Side effects:
- *      Sets the initial value for the uptime base.
+ *      Sets the initial values for the uptime state, and schedules
+ *      the uptime timer.
  *
  *----------------------------------------------------------------------
  */
@@ -1658,7 +1756,22 @@ HostIF_WaitForFreePages(unsigned int timeoutMs)  // IN:
 void
 HostIF_InitUptime(void)
 {
-   uptime_base = ktime_get_ns();
+   struct timeval tv;
+
+   uptimeState.jiffiesBase = jiffies;
+   do_gettimeofday(&tv);
+   Atomic_Write64(&uptimeState.uptimeBase,
+                  -(tv.tv_usec * (UPTIME_FREQ / 1000000) +
+                    tv.tv_sec * UPTIME_FREQ));
+
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 15, 0) && !defined(timer_setup)
+   init_timer(&uptimeState.timer);
+   uptimeState.timer.function = (void *)HostIFUptimeResyncMono;
+   uptimeState.timer.data = (unsigned long)&uptimeState.timer;
+#else
+   timer_setup(&uptimeState.timer, HostIFUptimeResyncMono, 0);
+#endif
+   mod_timer(&uptimeState.timer, jiffies + HZ);
 }
 
 
@@ -1667,13 +1780,13 @@ HostIF_InitUptime(void)
  *
  * HostIF_CleanupUptime --
  *
- *      No-op, left for backward compatibility.
+ *      Cleanup uptime state, called at module unloading time.
  *
  * Results:
  *      None
  *
  * Side effects:
- *      None
+ *      Deschedule the uptime timer.
  *
  *----------------------------------------------------------------------
  */
@@ -1681,6 +1794,7 @@ HostIF_InitUptime(void)
 void
 HostIF_CleanupUptime(void)
 {
+   del_timer_sync(&uptimeState.timer);
 }
 
 
@@ -1704,10 +1818,9 @@ HostIF_CleanupUptime(void)
 uint64
 HostIF_ReadUptime(void)
 {
-   u64 tm;
+   unsigned long jifs;
 
-   tm = ktime_get_ns();
-   return (tm - uptime_base) / (NSEC_PER_SEC / UPTIME_FREQ);
+   return HostIFReadUptimeWork(&jifs);
 }
 
 
@@ -2027,27 +2140,6 @@ 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, 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
-
 /*
  *----------------------------------------------------------------------
  *
@@ -2067,11 +2159,15 @@ static long compat_copy_from_kernel_nofault(void *dst, VA src, size_t size)
 static Bool
 isVAReadable(VA r)  // IN:
 {
+   mm_segment_t old_fs;
    uint32 dummy;
-   long ret;
+   int ret;
 
+   old_fs = get_fs();
+   set_fs(get_ds());
    r = APICR_TO_ADDR(r, APICR_VERSION);
-   ret = compat_copy_from_kernel_nofault(&dummy, r, sizeof(dummy));
+   ret = HostIF_CopyFromUser(&dummy, r, sizeof dummy);
+   set_fs(old_fs);
 
    return ret == 0;
 }
@@ -2101,7 +2197,7 @@ SetVMAPICAddr(VMDriver *vm, // IN/OUT: driver state
    volatile void *hostapic;
 
    ASSERT_ON_COMPILE(APICR_SIZE <= PAGE_SIZE);
-   hostapic = (volatile void *) ioremap(ma, PAGE_SIZE);
+   hostapic = (volatile void *) ioremap_nocache(ma, PAGE_SIZE);
    if (hostapic) {
       if ((APIC_VERSIONREG(hostapic) & 0xF0) == 0x10) {
          vm->hostAPIC.base = (volatile uint32 (*)[4]) hostapic;
@@ -2257,41 +2353,44 @@ 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;
    }
 
-   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;
+   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);
    }
-   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.
     */
-#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
+
+   res = file->f_op->read(file, (char *) &value, sizeof value, &file->f_pos);
+
    if (res == sizeof value) {
       res = MX_WAITNORMAL;
    } else {
@@ -2300,6 +2399,7 @@ HostIF_SemaphoreWait(VMDriver *vm,   // IN:
       }
    }
 
+   set_fs(old_fs);
    fput(file);
 
    /*
@@ -2382,8 +2482,8 @@ 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
@@ -2393,34 +2493,22 @@ HostIF_SemaphoreSignal(uint64 *args)  // IN:
       return MX_WAITERROR;
    }
 
-   /*
-    * 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;
-   }
+   old_fs = get_fs();
+   set_fs(get_ds());
 
    /*
     * 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.
     */
-#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
+
+   res = file->f_op->write(file, (char *) &value, sizeof value, &file->f_pos);
 
    if (res == sizeof value) {
       res = MX_WAITNORMAL;
    }
 
+   set_fs(old_fs);
    fput(file);
 
    /*
@@ -2939,9 +3027,12 @@ static int
 HostIFFastClockThread(void *unused)  // IN:
 {
    int res;
+   mm_segment_t oldFS;
    unsigned int rate = 0;
    unsigned int prevRate = 0;
 
+   oldFS = get_fs();
+   set_fs(KERNEL_DS);
    allow_signal(SIGKILL);
 
    while ((rate = linuxState.fastClockRate) > MIN_RATE) {
@@ -2964,6 +3055,8 @@ HostIFFastClockThread(void *unused)  // IN:
    }
 
  out:
+   set_fs(oldFS);
+
    /*
     * Do not exit thread until we are told to do so.
     */
@@ -3061,7 +3154,7 @@ HostIF_SetFastClockRate(unsigned int rate) // IN: Frequency in Hz.
       }
    } else {
       if (linuxState.fastClockThread) {
-         send_sig(SIGKILL, linuxState.fastClockThread, 1);
+         force_sig(SIGKILL, linuxState.fastClockThread);
          kthread_stop(linuxState.fastClockThread);
 
          linuxState.fastClockThread = NULL;
@@ -3107,7 +3200,7 @@ HostIF_MapUserMem(VA addr,                  // IN: User memory virtual address
 
    ASSERT(handle);
 
-   if (!write_access_ok(p, size)) {
+   if (!access_ok(VERIFY_WRITE, 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

@@ -107,7 +107,7 @@ prebuild:: ;
 postbuild:: ;
 
 $(DRIVER_KO): prebuild
-	$(MAKE) -C $(BUILD_DIR) M=$$PWD SRCROOT=$$PWD/$(SRCROOT) \
+	$(MAKE) -C $(BUILD_DIR) SUBDIRS=$$PWD SRCROOT=$$PWD/$(SRCROOT) \
 	  MODULEBUILDDIR=$(MODULEBUILDDIR) modules
 	$(MAKE) -C $$PWD SRCROOT=$$PWD/$(SRCROOT) \
 	  MODULEBUILDDIR=$(MODULEBUILDDIR) postbuild

vmnet-only/Makefile.kernel

@@ -32,6 +32,6 @@ $(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 .cache.mk \
+	rm -rf $(DRIVER).o $(DRIVER).mod.o $(DRIVER).mod.c $(DRIVER).ko \
 	       .tmp_versions Module.symvers Modules.symvers Module.markers \
 	       modules.order $($(DRIVER)-y) .*.cmd .*.o.flags

vmnet-only/bridge.c

@@ -66,7 +66,7 @@
 #endif
 
 #if LOGLEVEL >= 4
-static u64 vnetTime;
+static struct timeval vnetTime;
 #endif
 
 typedef struct VNetBridge VNetBridge;
@@ -693,7 +693,7 @@ VNetBridgeReceiveFromVNet(VNetJack        *this, // IN: jack
 
 	 netif_rx_ni(clone);
 #	 if LOGLEVEL >= 4
-	 vnetTime = ktime_get_ns();
+	 do_gettimeofday(&vnetTime);
 #	 endif
       }
    }
@@ -1490,11 +1490,12 @@ VNetBridgeReceiveFromDev(struct sk_buff *skb,         // IN: packet to receive
 
 #  if LOGLEVEL >= 4
    {
-      u64 now;
-
-      now = ktime_get_ns();
+      struct timeval now;
+      do_gettimeofday(&now);
       LOG(3, (KERN_DEBUG "bridge-%s: time %d\n",
-	      bridge->name, (int)((now - vnetTime) / NSEC_PER_USEC)));
+	      bridge->name,
+	      (int)((now.tv_sec * 1000000 + now.tv_usec)
+                    - (vnetTime.tv_sec * 1000000 + vnetTime.tv_usec))));
    }
 #  endif
 

vmnet-only/procfs.c

@@ -161,14 +161,6 @@ 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,
@@ -176,7 +168,6 @@ static struct file_operations fops = {
    .release = single_release,
 };
 #endif
-#endif
 
 
 /*

vmnet-only/userif.c

@@ -78,55 +78,11 @@ 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_frag(frag) kmap((frag)->page)
-#   define compat_kunmap_frag(page) kunmap((frag)->page)
+#   define compat_kmap(page) kmap(page)
+#   define compat_kunmap(page) kunmap(page)
 #else
-#   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);
-}
+#   define compat_kmap(page) kmap((page).p)
+#   define compat_kunmap(page) kunmap((page).p)
 #endif
 
 /*
@@ -186,7 +142,7 @@ VNetUserIfMapPtr(VA uAddr,        // IN: pointer to user memory
                  struct page **p, // OUT: locked page
                  void **ptr)      // OUT: kernel mapped pointer
 {
-   if (!write_access_ok((void *)uAddr, size) ||
+   if (!access_ok(VERIFY_WRITE, (void *)uAddr, size) ||
        (((uAddr + size - 1) & ~(PAGE_SIZE - 1)) !=
         (uAddr & ~(PAGE_SIZE - 1)))) {
       return -EINVAL;
@@ -581,6 +537,7 @@ 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;
@@ -593,27 +550,29 @@ VNetCsumCopyDatagram(const struct sk_buff *skb,	// IN: skb to copy
       return -EINVAL;
    }
 
-   csum = compat_csum_and_copy_to_user(skb->data + offset, curr, len);
-   if (!csum)
-	   return -EFAULT;
+   csum = csum_and_copy_to_user(skb->data + offset, curr, len, 0, &err);
+   if (err) {
+      return err;
+   }
    curr += len;
 
    for (frag = skb_shinfo(skb)->frags;
 	frag != skb_shinfo(skb)->frags + skb_shinfo(skb)->nr_frags;
 	frag++) {
-      if (compat_skb_frag_size(frag) > 0) {
+      if (frag->size > 0) {
 	 unsigned int tmpCsum;
 	 const void *vaddr;
 
-	 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);
+	 vaddr = compat_kmap(frag->page);
+	 tmpCsum = csum_and_copy_to_user(vaddr + frag->page_offset,
+					 curr, frag->size, 0, &err);
+	 compat_kunmap(frag->page);
 
-	 if (!tmpCsum)
-		 return -EFAULT;
+	 if (err) {
+	    return err;
+	 }
 	 csum = csum_block_add(csum, tmpCsum, curr - buf);
-	 curr += compat_skb_frag_size(frag);
+	 curr += frag->size;
       }
    }
 

vmnet-only/vm_device_version.h

@@ -35,8 +35,6 @@
 #endif
 #endif
 
-#include <linux/pci_ids.h>
-
 /* LSILogic 53C1030 Parallel SCSI controller
  * LSILogic SAS1068 SAS controller
  */
@@ -55,10 +53,7 @@
  *    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
@@ -75,11 +70,7 @@
 #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