summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
-rw-r--r--src/target/adi_v5_jtag.c40
1 files changed, 24 insertions, 16 deletions
diff --git a/src/target/adi_v5_jtag.c b/src/target/adi_v5_jtag.c
index 8731a1a0..48b4a7b8 100644
--- a/src/target/adi_v5_jtag.c
+++ b/src/target/adi_v5_jtag.c
@@ -191,22 +191,30 @@ static int jtagdp_transaction_endcheck(struct adiv5_dap *dap)
/* too expensive to call keep_alive() here */
-#if 0
- /* Danger!!!! BROKEN!!!! */
- adi_jtag_scan_inout_check_u32(dap, JTAG_DP_DPACC,
- DP_CTRL_STAT, DPAP_READ, 0, &ctrlstat);
- /* Danger!!!! BROKEN!!!! Why will jtag_execute_queue() fail here????
- R956 introduced the check on return value here and now Michael Schwingen reports
- that this code no longer works....
-
- https://lists.berlios.de/pipermail/openocd-development/2008-September/003107.html
- */
- if ((retval = jtag_execute_queue()) != ERROR_OK)
- {
- LOG_ERROR("BUG: Why does this fail the first time????");
- }
- /* Why??? second time it works??? */
-#endif
+ /* Here be dragons!
+ *
+ * It is easy to be in a JTAG clock range where the target
+ * is not operating in a stable fashion. This happens
+ * for a few reasons:
+ *
+ * - the user may construct a simple test case to try to see
+ * if a higher JTAG clock works to eke out more performance.
+ * This simple case may pass, but more complex situations can
+ * fail.
+ *
+ * - The mostly works JTAG clock rate and the complete failure
+ * JTAG clock rate may be as much as 2-4x apart. This seems
+ * to be especially true on RC oscillator driven parts.
+ *
+ * So: even if calling adi_jtag_scan_inout_check_u32() multiple
+ * times here seems to "make things better here", it is just
+ * hiding problems with too high a JTAG clock.
+ *
+ * Note that even if some parts have RCLK/RTCK, that doesn't
+ * mean that RCLK/RTCK is the *correct* rate to run the JTAG
+ * interface at, i.e. RCLK/RTCK rates can be "too high", especially
+ * before the RC oscillator phase is not yet complete.
+ */
/* Post CTRL/STAT read; discard any previous posted read value
* but collect its ACK status.