import Workstation 15.1.0 module sources

vmmon-only/common/phystrack.c

@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998,2014 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998,2014,2019 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
@@ -24,9 +24,8 @@
  *    Depending on configuration phystracker provides either 2-level or
  *    3-level structure to track whether page (specified by its MPN) is
  *    locked or no.  Linux uses 3-level structures with top limit of
- *    1TB (32bit) or 16TB (64bit).  Windows use 2-level structures
- *    ready to hold 128GB (32bit) or 2TB (64bit) of memory.  On Mac
- *    limit is 2-level 128GB.
+ *    1TB (32bit) or 16TB (64bit).  Windows and Mac use 2-level structures
+ *    ready to hold 128GB (32bit) or 2TB (64bit) of memory.
  *
  *    2-level phystracker is built on top of 3-level one by collapsing
  *    middle level.
@@ -50,9 +49,7 @@
 #define BYTES_PER_ENTRY      (PAGE_SIZE)
 #define PHYSTRACK_L3_ENTRIES (8 * BYTES_PER_ENTRY) /* 128MB */
 
-#if defined(WINNT_DDK)
-#define PHYSTRACK_L1_ENTRIES (PHYSTRACK_MAX_SUPPORTED_GB * 8)
-#elif defined(__linux__)
+#if defined(__linux__)
 #define PHYSTRACK_L2_ENTRIES (BYTES_PER_ENTRY / sizeof(void *)) /* 64GB or 128GB */
 /*
  * Currently MPN is 32 bits.  15 bits are in L3, 9 bits are in L2,
@@ -60,7 +57,7 @@
  */
 #define PHYSTRACK_L1_ENTRIES (256) /* 16TB. */
 #else
-#define PHYSTRACK_L1_ENTRIES ((128 + 4) * 8) /* 128 GB */
+#define PHYSTRACK_L1_ENTRIES (PHYSTRACK_MAX_SUPPORTED_GB * 8)
 #endif
 
 #ifndef PHYSTRACK_L2_ENTRIES

vmmon-only/common/phystrack.h

@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998,2019 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
@@ -29,7 +29,7 @@
 #define INCLUDE_ALLOW_VMCORE
 #include "includeCheck.h"
 
-#if defined(WINNT_DDK)
+#if !defined (__linux__)
 /* XXX: can be more efficient based on server vs. desktop and version of Windows */
 #define PHYSTRACK_MAX_SUPPORTED_GB (2048 + 4) /* 2 TB 64-bit W2k8 + 4 GB PCI */
 #endif

vmmon-only/common/task.c

@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998-2018 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998-2019 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
@@ -2909,6 +2909,18 @@ Task_Switch(VMDriver *vm,  // IN
    }
 
    if (crosspage->crosspageData.moduleCallType == MODULECALL_INTR) {
+
+      /*
+       * Newer versions of Window expect EFLAGS_AC to be set when handling an
+       * interupt - PR  2248661.
+       */
+      if ((flags & EFLAGS_AC) != 0) {
+         uintptr_t   curFlags;
+         SAVE_FLAGS(curFlags);
+         curFlags |= EFLAGS_AC;
+         RESTORE_FLAGS(curFlags);
+      }
+
       /*
        * Note we must do the RAISE_INTERRUPT before ever enabling
        * interrupts or bad things have happened (might want to know exactly

vmmon-only/include/iocontrols.h

@@ -144,7 +144,7 @@ PtrToVA64(void const *ptr) // IN
  * See bora/doc/vmcore details.
  */
 
-#define VMMON_VERSION           (360 << 16 | 0)
+#define VMMON_VERSION           (361 << 16 | 0)
 #define VMMON_VERSION_MAJOR(v)  ((uint32) (v) >> 16)
 #define VMMON_VERSION_MINOR(v)  ((uint16) (v))
 
@@ -276,6 +276,7 @@ enum IOCTLCmd {
 #endif
 
 #if defined __APPLE__
+   IOCTLCMD(GET_NUM_RESPONDING_CPUS),
    IOCTLCMD(ALLOC_LOW_PAGES),
    IOCTLCMD(INIT_DRIVER),
    IOCTLCMD(BLUEPILL),

vmmon-only/include/x86cpuid.h

@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998-2018 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998-2019 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
@@ -465,6 +465,9 @@ FLAG(   6,  0, ECX,  3,  1, ENERGY_PERF_BIAS,                  NO,   0, FALSE)
 #define CPUID_7_EDX_28 \
 FLAG(   7,  0, EDX, 28,  1, LEVEL7EDX_RSVD1,                    NO,  0, FALSE)
 
+/*    LEVEL, SUB-LEVEL, REG, POS, SIZE, NAME,             MON SUPP, HWV, CPL3 */
+#define CPUID_7_EDX_10 \
+FLAG(   7,  0, EDX, 10,  1, LEAF7_RSVD,                         NO,  0, FALSE)
 #define CPUID_7_EDX_31 \
 FLAG(   7,  0, EDX, 31,  1, LEVEL7EDX_RSVD2,                    NO,  0, FALSE)
 
@@ -519,6 +522,7 @@ FLAG(   7,  0, ECX, 22,  1, RDPID,                             NO,   0, TRUE)  \
 FLAG(   7,  0, ECX, 30,  1, SGX_LC,                            YES, 17, FALSE) \
 FLAG(   7,  0, EDX,  2,  1, AVX512QVNNIW,                      YES, 16, TRUE)  \
 FLAG(   7,  0, EDX,  3,  1, AVX512QFMAPS,                      YES, 16, TRUE)  \
+CPUID_7_EDX_10                                                                 \
 FLAG(   7,  0, EDX, 26,  1, IBRSIBPB,                          ANY,  9, FALSE) \
 FLAG(   7,  0, EDX, 27,  1, STIBP,                             YES,  9, FALSE) \
 CPUID_7_EDX_28                                                                 \
@@ -1432,6 +1436,8 @@ CPUIDCheck(int32 eaxIn, int32 eaxInCheck,
 #define CPUID_STEPPING_KABYLAKE_ES     0x8  // Kabylake S/H/U/Y ES
 #define CPUID_STEPPING_COFFEELAKE_A    0xA  // Coffeelake U/S/H
 #define CPUID_STEPPING_COFFEELAKE_B    0xB  // Coffeelake S/H
+#define CPUID_STEPPING_CASCADELAKE_A   0x5  // Cascade Lake A-step
+#define CPUID_STEPPING_CASCADELAKE_B1  0x7  // Cascade Lake B1-step
 
 #define CPUID_MODEL_PIII_07    7
 #define CPUID_MODEL_PIII_08    8
@@ -1656,6 +1662,15 @@ CPUID_MODEL_IS_HASWELL(uint32 v) // IN: %eax from CPUID with %eax=1.
            effectiveModel == CPUID_MODEL_HASWELL_46);
 }
 
+static INLINE Bool
+CPUID_MODEL_IS_CASCADELAKE(uint32 v) // IN: %eax from CPUID with %eax=1.
+{
+   /* Assumes the CPU manufacturer is Intel. */
+   return CPUID_FAMILY_IS_P6(v) &&
+          CPUID_EFFECTIVE_MODEL(v) == CPUID_MODEL_SKYLAKE_55 &&
+          CPUID_EFFECTIVE_STEPPING(v) >= CPUID_STEPPING_CASCADELAKE_A &&
+          CPUID_EFFECTIVE_STEPPING(v) <= CPUID_STEPPING_CASCADELAKE_B1;
+}
 
 static INLINE Bool
 CPUID_MODEL_IS_SKYLAKE(uint32 v) // IN: %eax from CPUID with %eax=1.

vmmon-only/include/x86msr.h

@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 1998-2018 VMware, Inc. All rights reserved.
+ * Copyright (C) 1998-2019 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
@@ -120,6 +120,8 @@ MSRQuery;
 #define MSR_ARCH_CAPABILITIES_RDCL_NO             (1ULL << 0)
 #define MSR_ARCH_CAPABILITIES_IBRS_ALL            (1ULL << 1)
 #define MSR_ARCH_CAPABILITIES_RSBA                (1ULL << 2)
+#define MSR_ARCH_CAPABILITIES_NOL1F_VMENTRY       (1ULL << 3)
+#define MSR_ARCH_CAPABILITIES_SSB_NO              (1ULL << 4)
 
 
 #define MSR_SPEC_CTRL_IBRS                        (1UL << 0)

vmmon-only/linux/hostif.c

@@ -1638,9 +1638,13 @@ HostIF_EstimateLockedPageLimit(const VMDriver* vm,                // IN
     * since at least 2.6.0.
     */
 
+#if LINUX_VERSION_CODE < KERNEL_VERSION(5, 0, 0)
    extern unsigned long totalram_pages;
 
    unsigned int totalPhysicalPages = totalram_pages;
+#else
+   unsigned int totalPhysicalPages = totalram_pages();
+#endif
 
    /*
     * Use the memory information linux exports as of late for a more
@@ -1738,6 +1742,49 @@ HostIF_WaitForFreePages(unsigned int timeoutMs)  // IN:
 }
 
 
+/*
+ *----------------------------------------------------------------------
+ *
+ * HostIFGetTime --
+ *
+ *      Reads the current time in UPTIME_FREQ units.
+ *
+ * Results:
+ *      The uptime, in units of UPTIME_FREQ.
+ *
+ * Side effects:
+ *      None.
+ *
+ *----------------------------------------------------------------------
+ */
+
+static uint64
+HostIFGetTime(void)
+{
+#if LINUX_VERSION_CODE < KERNEL_VERSION(5, 0, 0)
+   struct timeval tv;
+
+   do_gettimeofday(&tv);
+   return tv.tv_usec * (UPTIME_FREQ / 1000000) + tv.tv_sec * UPTIME_FREQ;
+#else
+   struct timespec64 now;
+
+   /*
+    * Use raw time used by Posix timers.  This time is not affected by
+    * NTP adjustments, so it may drift from real time and monotonic time,
+    * but it will stay in sync with other timers.
+    */
+   ktime_get_raw_ts64(&now);
+   /*
+    * UPTIME_FREQ resolution is lower than tv_nsec,
+    * so we have to do division...
+    */
+   ASSERT_ON_COMPILE(1000000000 % UPTIME_FREQ == 0);
+   return now.tv_nsec / (1000000000 / UPTIME_FREQ) + now.tv_sec * UPTIME_FREQ;
+#endif
+}
+
+
 /*
  *----------------------------------------------------------------------
  *
@@ -1769,7 +1816,6 @@ HostIF_WaitForFreePages(unsigned int timeoutMs)  // IN:
 static uint64
 HostIFReadUptimeWork(unsigned long *j)  // OUT: current jiffies
 {
-   struct timeval tv;
    uint64 monotime, uptime, upBase, monoBase;
    int64 diff;
    uint32 version;
@@ -1784,13 +1830,12 @@ HostIFReadUptimeWork(unsigned long *j)  // OUT: current jiffies
       monoBase = uptimeState.monotimeBase;
    } while (!VersionedAtomic_EndTryRead(&uptimeState.version, version));
 
-   do_gettimeofday(&tv);
+   uptime = HostIFGetTime();
    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;
 
    /*
@@ -1895,13 +1940,11 @@ HostIFUptimeResyncMono(struct timer_list *timer)  // IN: ignored
 void
 HostIF_InitUptime(void)
 {
-   struct timeval tv;
+   uint64 tm;
 
    uptimeState.jiffiesBase = jiffies;
-   do_gettimeofday(&tv);
-   Atomic_Write64(&uptimeState.uptimeBase,
-                  -(tv.tv_usec * (UPTIME_FREQ / 1000000) +
-                    tv.tv_sec * UPTIME_FREQ));
+   tm = HostIFGetTime();
+   Atomic_Write64(&uptimeState.uptimeBase, -tm);
 
 #if LINUX_VERSION_CODE < KERNEL_VERSION(4, 15, 0) && !defined(timer_setup)
    init_timer(&uptimeState.timer);
@@ -3405,7 +3448,12 @@ HostIF_MapUserMem(VA addr,                  // IN: User memory virtual address
 
    ASSERT(handle);
 
-   if (!access_ok(VERIFY_WRITE, p, size)) {
+#if LINUX_VERSION_CODE < KERNEL_VERSION(5, 0, 0)
+   if (!access_ok(VERIFY_WRITE, p, size))
+#else
+   if (!access_ok(p, size))
+#endif
+   {
       printk(KERN_ERR "%s: Couldn't verify write to uva 0x%p with size %"
              FMTSZ"u\n", __func__, p, size);
 

vmnet-only/userif.c

@@ -137,16 +137,21 @@ UserifLockPage(VA addr) // IN
  */
 
 static INLINE int
-VNetUserIfMapPtr(VA uAddr,        // IN: pointer to user memory
+VNetUserIfMapPtr(VA64 uAddr,      // IN: pointer to user memory
                  size_t size,     // IN: size of data
                  struct page **p, // OUT: locked page
                  void **ptr)      // OUT: kernel mapped pointer
 {
-   if (!access_ok(VERIFY_WRITE, (void *)uAddr, size) ||
-       (((uAddr + size - 1) & ~(PAGE_SIZE - 1)) !=
-        (uAddr & ~(PAGE_SIZE - 1)))) {
+   uint8 v;
+
+   /* Check area does not straddle two pages. */
+   if ((uAddr & (PAGE_SIZE - 1)) + size > PAGE_SIZE) {
       return -EINVAL;
    }
+   /* Check if it is user's area.  UserifLockPage() checks writability. */
+   if (copy_from_user(&v, (void *)(unsigned long)uAddr, sizeof v) != 0) {
+      return -EFAULT;
+   }
 
    *p = UserifLockPage(uAddr);
    if (*p == NULL) {
@@ -158,7 +163,7 @@ VNetUserIfMapPtr(VA uAddr,        // IN: pointer to user memory
 }
 
 static INLINE int
-VNetUserIfMapUint32Ptr(VA uAddr,        // IN: pointer to user memory
+VNetUserIfMapUint32Ptr(VA64 uAddr,      // IN: pointer to user memory
                        struct page **p, // OUT: locked page
                        uint32 **ptr)    // OUT: kernel mapped pointer
 {
@@ -201,7 +206,7 @@ VNetUserIfSetupNotify(VNetUserIF *userIf, // IN
       return -EBUSY;
    }
 
-   if ((retval = VNetUserIfMapUint32Ptr((VA)vn->pollPtr, &pollPage,
+   if ((retval = VNetUserIfMapUint32Ptr(vn->pollPtr, &pollPage,
                                         &pollPtr)) < 0) {
       return retval;
    }
@@ -213,7 +218,7 @@ VNetUserIfSetupNotify(VNetUserIF *userIf, // IN
       goto error_free;
    }
 
-   if ((retval = VNetUserIfMapUint32Ptr((VA)vn->recvClusterPtr,
+   if ((retval = VNetUserIfMapUint32Ptr(vn->recvClusterPtr,
                                         &recvClusterPage,
                                         &recvClusterCount)) < 0) {
       goto error_free;