serial: reset thri_pending on IER writes with THRI=0

This is responsible for failure of migration from 2.2 to 2.1, because
thr_ipending is always one in practice.

serial.c is setting thr_ipending unconditionally.  However, thr_ipending
is not used at all if THRI=0, and it will be overwritten again the next
time THRE or THRI changes.  For that reason, we can set thr_ipending to
zero every time THRI is reset.

There is disagreement on whether LSR.THRE should be resampled when IER.THRI
goes from 1 to 1.  This patch does not touch the code, leaving that for
QEMU 2.3+.

This has no semantic change and is enough to fix migration in the common
case where the interrupt is not pending or is reported in IIR.  It does not
change the migration format, so 2.2.0 -> 2.1 will remain broken but we
can fix 2.2.1 -> 2.1 without breaking 2.2.1 <-> 2.2.0.

The case that remains broken (the one in which the subsection is strictly
necessary) is when THRE=1, the THRI interrupt has *not* been acknowledged
yet, and a higher-priority interrupt comes.  In this case, you need the
subsection to tell the source that the lower-priority THRI interrupt is
pending.  The subsection's breakage of migration, in this case, prevents
continuing the VM on the destination with an invalid state.

Cc: qemu-stable@nongnu.org
Reported-by: Igor Mammedov <imammedo@redhat.com>
Reviewed-by: Dr. David Alan Gilbert <dgilbert@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

hw/char/serial.c

@@ -350,10 +350,24 @@ static void serial_ioport_write(void *opaque, hwaddr addr, uint64_t val,
                      s->poll_msl = 0;
                 }
             }
-            if (s->lsr & UART_LSR_THRE) {
+
+            /* Turning on the THRE interrupt on IER can trigger the interrupt
+             * if LSR.THRE=1, even if it had been masked before by reading IIR.
+             * This is not in the datasheet, but Windows relies on it.  It is
+             * unclear if THRE has to be resampled every time THRI becomes
+             * 1, or only on the rising edge.  Bochs does the latter, and Windows
+             * always toggles IER to all zeroes and back to all ones.  But for
+             * now leave it as it has always been in QEMU.
+             *
+             * If IER.THRI is zero, thr_ipending is not used.  Set it to zero
+             * so that the thr_ipending subsection is not migrated.
+             */
+            if ((s->ier & UART_IER_THRI) && (s->lsr & UART_LSR_THRE)) {
                 s->thr_ipending = 1;
-                serial_update_irq(s);
+            } else {
+                s->thr_ipending = 0;
             }
+            serial_update_irq(s);
         }
         break;
     case 2: