MFC r274821:

Merge from CheriBSD:

commit d0c7d235c09fc65dbdb278e7016a96f79c6a49cc
  Make the Altera JTAG UART device driver slightly more forgiving of
  the foibles of a sub-par hrdware interface by increasing the timeout
  for spotting JTAG polling from one to two seconds.

commit 19ed45a18832560dab967c179d83b71081c3a220
  Update comment.

commit 8edfe803f033cc8e33229f99894c2b7496a44d5f
  Add a comment about a device-driver race condition that could cause the BERI
  pipeline to wedge awaiting JTAG in the event that both the low-level console
  and the tty layer decide to write to the JTAG FIFO just before JTAG is
  disconnected.  Resolving this race is a bit tricky as it looks like there
  isn't a way to 'give the character back' to the tty layer when we discover
  the race.  The easy fix is to drop the character, which we don't yet do, but
  perhaps should as that is a better outcome than wedging the pipeline.

commit 2ea26cf579c9defcf31e413e7c9b0fbc159237fc
  Add a comment about an inherent race with hardware in the Altera JTAG
  UART's low-level console code.

Submitted by:	rwatson
Sponsored by:	DARPA, AFRL

2 files changed, 21 insertions, 2 deletions

diff --git a/sys/dev/altera/jtag_uart/altera_jtag_uart_cons.c b/sys/dev/altera/jtag_uart/altera_jtag_uart_cons.c
index ae84597..96d20ce 100644
--- a/sys/dev/altera/jtag_uart/altera_jtag_uart_cons.c
+++ b/sys/dev/altera/jtag_uart/altera_jtag_uart_cons.c
@@ -221,6 +221,9 @@ aju_cons_write(char ch)
 	 * disconnection.
 	 *
 	 * XXXRW: The polling delay may require tuning.
+	 *
+	 * XXXRW: Notice the inherent race with hardware: in clearing the
+	 * bit, we may race with hardware setting the same bit.
 	 */
 	v = aju_cons_control_read();
 	if (v & ALTERA_JTAG_UART_CONTROL_AC) {
diff --git a/sys/dev/altera/jtag_uart/altera_jtag_uart_tty.c b/sys/dev/altera/jtag_uart/altera_jtag_uart_tty.c
index ad5b8a4..dd46ea0 100644
--- a/sys/dev/altera/jtag_uart/altera_jtag_uart_tty.c
+++ b/sys/dev/altera/jtag_uart/altera_jtag_uart_tty.c
@@ -65,9 +65,9 @@ static struct ttydevsw aju_ttydevsw = {
 
 /*
  * When polling for the AC bit, the number of times we have to not see it
- * before assuming JTAG has disappeared on us.  By default, one second.
+ * before assuming JTAG has disappeared on us.  By default, two seconds.
  */
-#define	AJU_JTAG_MAXMISS		5
+#define	AJU_JTAG_MAXMISS		10
 
 /*
  * Polling intervals for input/output and JTAG connection events.
@@ -255,6 +255,22 @@ aju_handle_output(struct altera_jtag_uart_softc *sc, struct tty *tp)
 			if (ttydisc_getc(tp, &ch, sizeof(ch)) != sizeof(ch))
 				panic("%s: ttydisc_getc", __func__);
 			AJU_LOCK(sc);
+
+			/*
+			 * XXXRW: There is a slight race here in which we test
+			 * for writability, drop the lock, get the character
+			 * from the tty layer, re-acquire the lock, and then
+			 * write.  It's possible for other code --
+			 * specifically, the low-level console -- to have
+			 * written in the mean time, which might mean that
+			 * there is no longer space.  The BERI memory bus will
+			 * cause this write to block, wedging the processor
+			 * until space is available -- which could be a while
+			 * if JTAG is not attached!
+			 *
+			 * The 'easy' fix is to drop the character if WSPACE
+			 * has become unset.  Not sure what the 'hard' fix is.
+			 */
 			aju_data_write(sc, ch);
 		} else {
 			/*