[src/jdolecek-ncqfixes]: src/sys/dev/ata relax the NOERROR + C_CHAOS, just re...

[jdolecek <jdolecek%NetBSD.org@localhost>]

details:   https://anonhg.NetBSD.org/src/rev/42012adbdda7
branches:  jdolecek-ncqfixes
changeset: 1025088:42012adbdda7
user:      jdolecek <jdolecek%NetBSD.org@localhost>
date:      Thu Oct 04 19:42:01 2018 +0000

description:
relax the NOERROR + C_CHAOS, just rerun the xfer instead of panic, to handle
another variant of bad/missing AHCI error recovery, this time under Parallels

diffstat:

 sys/dev/ata/TODO.ncq |   5 +++--
 sys/dev/ata/wd.c     |  20 +++++++++++++++-----
 2 files changed, 18 insertions(+), 7 deletions(-)

diffs (58 lines):

diff -r 58b4111c7dfd -r 42012adbdda7 sys/dev/ata/TODO.ncq
--- a/sys/dev/ata/TODO.ncq      Thu Oct 04 17:59:35 2018 +0000
+++ b/sys/dev/ata/TODO.ncq      Thu Oct 04 19:42:01 2018 +0000
@@ -1,6 +1,7 @@
 jdolecek-ncqfixes goals:
-- fix ahci(4) error handling under paralles - invalid bio via WD_CHAOS_MONKEY
-  ends up being handled as NOERROR, triggering KASSERT() in wd(4)
+- re-check READ LOG EXT handling under native and Parallels to make sure
+  the NOERROR under Parallels is their bug and not ours
+- run recovery via atathread, move to new function and share ahci/siisata/mvsata
 - remove controller-specific slot bitmaps (ic/siisata.c, ic/ahcisata_core.c)
 
 Bugs
diff -r 58b4111c7dfd -r 42012adbdda7 sys/dev/ata/wd.c
--- a/sys/dev/ata/wd.c  Thu Oct 04 17:59:35 2018 +0000
+++ b/sys/dev/ata/wd.c  Thu Oct 04 19:42:01 2018 +0000
@@ -1,4 +1,4 @@
-/*     $NetBSD: wd.c,v 1.441.2.8 2018/10/04 17:53:23 jdolecek Exp $ */
+/*     $NetBSD: wd.c,v 1.441.2.9 2018/10/04 19:42:01 jdolecek Exp $ */
 
 /*
  * Copyright (c) 1998, 2001 Manuel Bouyer.  All rights reserved.
@@ -54,7 +54,7 @@
  */
 
 #include <sys/cdefs.h>
-__KERNEL_RCSID(0, "$NetBSD: wd.c,v 1.441.2.8 2018/10/04 17:53:23 jdolecek Exp $");
+__KERNEL_RCSID(0, "$NetBSD: wd.c,v 1.441.2.9 2018/10/04 19:42:01 jdolecek Exp $");
 
 #include "opt_ata.h"
 #include "opt_wd.h"
@@ -914,13 +914,23 @@
                bp->b_error = EIO;
                break;
        case NOERROR:
+#ifdef WD_CHAOS_MONKEY
+               /*
+                * For example Parallels AHCI emulation doesn't actually
+                * return error for the invalid I/O, so just re-run
+                * the request and do not panic.
+                */
+               if (__predict_false(xfer->c_flags & C_CHAOS)) {
+                       xfer->c_bio.error = REQUEUE;
+                       errmsg = "chaos noerror";
+                       goto retry2;
+               }
+#endif
+
 noerror:       if ((xfer->c_bio.flags & ATA_CORR) || xfer->c_retries > 0)
                        device_printf(dksc->sc_dev,
                            "soft error (corrected) xfer %"PRIxPTR"\n",
                            (intptr_t)xfer & PAGE_MASK);
-#ifdef WD_CHAOS_MONKEY
-               KASSERT((xfer->c_flags & C_CHAOS) == 0);
-#endif
                break;
        case ERR_NODEV:
                bp->b_error = EIO;