From cb0de21d0cb6b899be30b6ce9c48d93f75a6c345 Mon Sep 17 00:00:00 2001 From: =?utf8?q?=C3=98yvind=20Harboe?= Date: Tue, 21 Sep 2010 22:13:09 +0200 Subject: [PATCH] jtagdp: remove #if 0'd kludges and explain why the code is correct MIME-Version: 1.0 Content-Type: text/plain; charset=utf8 Content-Transfer-Encoding: 8bit short story: if the JTAG clock is too high, then the behavior will be flaky and kludging the code may seem to make things beter, but really it's just a red herring. Signed-off-by: Øyvind Harboe --- src/target/adi_v5_jtag.c | 40 ++++++++++++++++++++++++---------------- 1 file changed, 24 insertions(+), 16 deletions(-) diff --git a/src/target/adi_v5_jtag.c b/src/target/adi_v5_jtag.c index 8731a1a09..48b4a7b8a 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. -- 2.11.4.GIT