vmmon: include <asm/irq_vectors.h> explicitly

After mainline commit 13c01139b171 ("x86/headers: Remove APIC headers from
<asm/smp.h>") in 5.9-rc1, APIC headers are no longer included via
<asm/smp.h> so that linux/hostif.c will use incorrect fallback definitions
of SPURIOUS_APIC_VECTOR, POSTED_INTR_VECTOR and ERROR_APIC_VECTOR even if
built against kernel where these are defined.

Include <asm/irq_vectors.h> in linux/hostif.c explicitly to avoid that.

vmmon-only/common/crosspage.c

@@ -47,11 +47,9 @@
 
 #ifdef __linux__
 #   include "driver-config.h"
-/*
- * linux/frame.h dates back to 4.6-rc1, we need the STACK_FRAME_NON_STANDARD
- * definition from it which is there from the start.
- */
-#   if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 6, 0)
+#   if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 10, 0)
+#      include <linux/objtool.h>
+#   elif LINUX_VERSION_CODE >= KERNEL_VERSION(4, 6, 0)
 #      include <linux/frame.h>
 #   endif
 #endif

vmmon-only/common/hostif.h

@@ -110,7 +110,7 @@ EXTERN int   HostIF_SemaphoreWait(VMDriver *vm,
                                   Vcpuid vcpuid,
                                   uint64 *args);
 
-EXTERN int   HostIF_SemaphoreSignal(uint64 *args);
+EXTERN int   HostIF_SemaphoreSignal(VMDriver *vm, uint64 *args);
 
 EXTERN void  HostIF_SemaphoreForceWakeup(VMDriver *vm, const VCPUSet *vcs);
 EXTERN void  HostIF_IPI(VMDriver *vm, const VCPUSet *vcs);

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/iocontrols.h

@@ -148,7 +148,7 @@ PtrToVA64(void const *ptr) // IN
  * the NT specific VMX86_DRIVER_VERSION.
  */
 
-#define VMMON_VERSION           (398 << 16 | 0)
+#define VMMON_VERSION           (401 << 16 | 0)
 #define VMMON_VERSION_MAJOR(v)  ((uint32) (v) >> 16)
 #define VMMON_VERSION_MINOR(v)  ((uint16) (v))
 
@@ -533,6 +533,8 @@ typedef struct IPIVectors {
  */
 typedef struct VMCreateBlock {
    VA64               bsBlob;        // IN: User VA of the VMM bootstrap blob.
+   VA64               vmmonData;     // IN: User VA of a userlevel scratch area
+                                     //     required by the Linux vmmon
    uint32             bsBlobSize;    // IN: Size of VMM bootstrap blob.
    uint32             numVCPUs;      // IN: Number of VCPUs.
    uint16             vmid;          // OUT: VM ID for the created VM.

vmmon-only/include/modulecall.h

@@ -421,9 +421,7 @@ VMCrossPageData;
 
 #define MX_WAITINTERRUPTED     3
 #define MX_WAITTIMEDOUT        2
-#define MX_WAITNORMAL          1  // Must equal one; see linux module code.
-#define MX_WAITERROR           0  // Use MX_ISWAITERROR() to test for error.
+#define MX_WAITNORMAL          1
+#define MX_WAITERROR           0
 
-// Any zero or negative value denotes error.
-#define MX_ISWAITERROR(e)      ((e) <= MX_WAITERROR)
 #endif

vmmon-only/include/vcpuset.h

@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 2002-2019 VMware, Inc. All rights reserved.
+ * Copyright (C) 2002-2020 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
@@ -44,10 +44,11 @@
 #   include "str.h"     /* Str_Snprintf */
 #   define VCS_SNPRINTF Str_Snprintf
 #elif defined MONITOR_APP
-#   include <stdio.h>   /* libc snprintf */
 #   if defined WIN32
-#      define VCS_SNPRINTF _snprintf
+#      include "str.h"     /* Str_Snprintf */
+#      define VCS_SNPRINTF Str_Snprintf
 #   else
+#      include <stdio.h>   /* libc snprintf */
 #      define VCS_SNPRINTF snprintf
 #   endif
 #elif defined VMM || defined VMKERNEL

vmmon-only/linux/driver.c

@@ -799,6 +799,7 @@ LinuxDriver_Ioctl(struct file *filp,    // IN:
       } else {
          device->vm = vm;
          args.vmid = vm->userID;
+         vm->vmhost->vmmonData = (void *__user)args.vmmonData;
          retval = HostIF_CopyToUser((VA64)ioarg, &args, sizeof args);
       }
       break;

vmmon-only/linux/hostif.c

@@ -77,7 +77,6 @@
 
 #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
@@ -2349,16 +2348,23 @@ HostIF_VMLockIsHeld(VMDriver *vm) // IN
 static Bool
 isVAReadable(VA r)  // IN:
 {
-   mm_segment_t old_fs;
    uint32 dummy;
    int ret;
 
+   r = APICR_TO_ADDR(r, APICR_VERSION);
+#ifdef HAVE_GET_KERNEL_NOFAULT
+   ret = get_kernel_nofault(dummy, (void *)r);
+#else
+   {
+   mm_segment_t old_fs;
+
    old_fs = get_fs();
    set_fs(KERNEL_DS);
-   r = APICR_TO_ADDR(r, APICR_VERSION);
+
    ret = HostIF_CopyFromUser(&dummy, r, sizeof dummy);
    set_fs(old_fs);
-
+   }
+#endif
    return ret == 0;
 }
 #endif
@@ -2524,8 +2530,10 @@ HostIF_GetPerfCtrVector(void)
  *    Perform the semaphore wait (P) operation, possibly blocking.
  *
  * Result:
- *    1 (which equals MX_WAITNORMAL) if success,
- *    negated error code otherwise.
+ *      On success: MX_WAITINTERRUPTED (3) if interrupted by a Unix signal.
+ *                  MX_WAITTIMEDOUT (2) if the timeout occurred.
+ *                  MX_WAITNORMAL (1) if the event was raised.
+ *      On error: MX_WAITERROR (0) on generic error.
  *
  * Side-effects:
  *    None
@@ -2552,10 +2560,15 @@ HostIF_SemaphoreWait(VMDriver *vm,   // IN:
    if (file == NULL) {
       return MX_WAITERROR;
    }
+   if (file->f_op->poll == NULL) {
+      fput(file);
+      Warning("%s: File poll operation is NULL\n", __func__);
+      return MX_WAITERROR;
+   }
 
    poll_initwait(&table);
    current->state = TASK_INTERRUPTIBLE;
-   mask = compat_vfs_poll(file, &table.pt);
+   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
@@ -2576,18 +2589,17 @@ HostIF_SemaphoreWait(VMDriver *vm,   // IN:
 #else
    res = kernel_read(file, &value, sizeof value, &file->f_pos);
 #endif
-   if (res == sizeof value) {
-      res = MX_WAITNORMAL;
-   } else {
-      if (res == 0) {
-         res = -EBADF;
-      }
-   }
-
    fput(file);
 
+   if (res == sizeof value) {
+      return MX_WAITNORMAL;
+   }
+   if (res == 0) {
+      res = -EBADF;
+   }
+
    /*
-    * Handle benign errors:
+    * Handle various errors:
     * EAGAIN is MX_WAITTIMEDOUT.
     * The signal-related errors are all mapped into MX_WAITINTERRUPTED.
     */
@@ -2606,6 +2618,10 @@ HostIF_SemaphoreWait(VMDriver *vm,   // IN:
    case -EBADF:
       res = MX_WAITERROR;
       break;
+   default:
+      Warning("Unexpected error: %s err=%d\n", __FUNCTION__, res);
+      res = MX_WAITERROR;
+      break;
    }
    return res;
 }
@@ -2659,7 +2675,6 @@ HostIF_SemaphoreForceWakeup(VMDriver *vm,       // IN:
  *      On error: MX_WAITINTERRUPTED (3) if interrupted by a Unix signal (we
  *                   can block on a preemptive kernel).
  *                MX_WAITERROR (0) on generic error.
- *                Negated system error (< 0).
  *
  * Side-effects:
  *      None
@@ -2668,57 +2683,64 @@ HostIF_SemaphoreForceWakeup(VMDriver *vm,       // IN:
  */
 
 int
-HostIF_SemaphoreSignal(uint64 *args)  // IN:
+HostIF_SemaphoreSignal(VMDriver *vm,  // IN:
+                       uint64 *args)  // IN:
 {
-   struct file *file;
    int res;
    int signalFD = args[1];
-   uint64 value = 1;  // make an eventfd happy should it be there
+   struct file *file;
 
    file = vmware_fget(signalFD);
-   if (!file) {
-      return MX_WAITERROR;
+   if (file == NULL) {
+      return FALSE;
    }
 
    /*
     * 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
-
-   if (res == sizeof value) {
-      res = MX_WAITNORMAL;
+   if (file->f_op->write != NULL) {
+      /*
+       * PR2650348 - The kernel_write function signature changed in the vanilla
+       * kernel in 4.14.0 and the signature update was backported before that
+       * point in a kernel used by OpenSUSE 15.1 so call the function directly
+       * instead of via the wrapper.
+       *
+       * Kernel 5.10 removed the set_fs functionality which means a user pointer
+       * must be passed into the write function or a -EFAULT will be returned.
+       */
+      res = file->f_op->write(file, vm->vmhost->vmmonData, sizeof (uint64),
+                              &file->f_pos);
+   } else {
+      Warning("%s: File write operation is NULL\n", __func__);
+      res = -ENOTSUPP;
    }
-
    fput(file);
 
+   if (res == sizeof (uint64)) {
+      return MX_WAITNORMAL;
+   }
+
    /*
-    * Handle benign errors:
+    * Handle various errors:
     * EAGAIN is MX_WAITTIMEDOUT.
     * The signal-related errors are all mapped into MX_WAITINTERRUPTED.
     */
-
    switch (res) {
    case -EAGAIN:
-      // The pipe is full, so it is already signalled. Success.
-      res = MX_WAITNORMAL;
-      break;
+      /* The pipe is full, so it is already signalled. Success. */
+      return MX_WAITNORMAL;
    case -EINTR:
    case -ERESTART:
    case -ERESTARTSYS:
    case -ERESTARTNOINTR:
    case -ERESTARTNOHAND:
-      res = MX_WAITINTERRUPTED;
-      break;
+      return MX_WAITINTERRUPTED;
+   default:
+      Warning("Unexpected error: %s err=%d\n", __FUNCTION__, res);
+      return MX_WAITERROR;
    }
-   return res;
 }
 
 
@@ -3243,12 +3265,9 @@ 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) {
@@ -3271,8 +3290,6 @@ HostIFFastClockThread(void *unused)  // IN:
    }
 
  out:
-   set_fs(oldFS);
-
    /*
     * Do not exit thread until we are told to do so.
     */

vmmon-only/linux/vmhost.h

@@ -92,6 +92,11 @@ typedef struct VMHost {
     * as pages for "AWE" guest memory.
     */
    struct PhysTracker  *AWEPages;
+   /*
+    * Pointer to a userlevel 64-bit area containing the value 1.
+    * This is used for HostIF_SemaphoreSignal.
+    */
+   void *__user vmmonData;
 } VMHost;
 
 #endif

vmmon-only/vmcore/moduleloop.c

@@ -180,7 +180,7 @@ skipTaskSwitch:;
       }
 
       case MODULECALL_SEMASIGNAL: {
-         retval = HostIF_SemaphoreSignal(crosspage->args);
+         retval = HostIF_SemaphoreSignal(vm, crosspage->args);
 
          if (retval == MX_WAITINTERRUPTED) {
              crosspage->moduleCallInterrupted = TRUE;
@@ -362,7 +362,13 @@ skipTaskSwitch:;
       default:
          Warning("ModuleCall %d not supported\n", crosspage->moduleCallType);
       }
-      ASSERT(retval == (uint64)((uint32)retval));
+
+      if (retval != (uint64)((uint32)retval)) {
+         Warning("Unexpected error in modulecall %u (%llu)\n",
+                 crosspage->moduleCallType, retval);
+         bailValue = USERCALL_SWITCHERR;
+         goto bailOut;
+      }
       crosspage->retval = (uint32)retval;
 #if defined(__linux__)
       cond_resched(); // Other kernels are preemptible

vmnet-only/smac.c

@@ -1,5 +1,5 @@
 /*********************************************************
- * Copyright (C) 2002,2018 VMware, Inc. All rights reserved.
+ * Copyright (C) 2002,2018,2020 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
@@ -4327,7 +4327,7 @@ LookupTypeClass(unsigned short typeValue) // IN: ethernet type
  *----------------------------------------------------------------------
  */
 
-#if _WIN32
+#ifdef _WIN32
 _Use_decl_annotations_
 #endif
 void

vmnet-only/userif.c

@@ -561,7 +561,12 @@ VNetCsumCopyDatagram(const struct sk_buff *skb,	// IN: skb to copy
       return -EINVAL;
    }
 
+#if COMPAT_LINUX_VERSION_CHECK_LT(5, 10, 0)
    csum = csum_and_copy_to_user(skb->data + offset, curr, len, 0, &err);
+#else
+   csum = csum_and_copy_to_user(skb->data + offset, curr, len);
+   err = (csum == 0) ? -EFAULT : 0;
+#endif
    if (err) {
       return err;
    }
@@ -575,8 +580,14 @@ VNetCsumCopyDatagram(const struct sk_buff *skb,	// IN: skb to copy
 	 const void *vaddr;
 
 	 vaddr = kmap(skb_frag_page(frag));
+#if COMPAT_LINUX_VERSION_CHECK_LT(5, 10, 0)
 	 tmpCsum = csum_and_copy_to_user(vaddr + skb_frag_off(frag),
 					 curr, skb_frag_size(frag), 0, &err);
+#else
+	 tmpCsum = csum_and_copy_to_user(vaddr + skb_frag_off(frag),
+					 curr, skb_frag_size(frag));
+         err = (tmpCsum == 0) ? -EFAULT : 0;
+#endif
 	 kunmap(skb_frag_page(frag));
 
 	 if (err) {