Current problem reports assigned to you

[Barry Bouwsma]

On Tue, 11 Jan 2005 11:46:16 -0800, Julian Elischer wrote:

> >| a [2003/03/02] kern/48849  usb         Maxtor XT5000 causes panic in boot
> >I have such a device, and the only problem I've had (4.x) is minor,
> >for which I added the NO_GETMAXLUN quirk.  I use it all the time.

> got a patch?

Yes, but it doesn't address the problem -- if it exists -- that prompted
the PR.

-- cut'n'paste part of diff - tab damage -- danger --
--- /stand/FreeBSD4-src/source-hacks/sys/dev/usb/umass.c-NEWER-current  Mon Dec
27 19:28:28 2004
+++ /stand/FreeBSD4-src/source-hacks/sys/dev/usb/umass.c        Tue Dec 28 21:07
:39 2004
@@ -369,6 +384,12 @@
          UMASS_PROTO_SCSI | UMASS_PROTO_BBB,
          NO_QUIRKS
        },
+/* XXX HACK */
+       { USB_VENDOR_MAXTOR, USB_PRODUCT_MAXTOR_5000, RID_WILDCARD,
+         UMASS_PROTO_SCSI | UMASS_PROTO_BBB,
+         NO_GETMAXLUN
+       },
+/* XXX HACK */
        { USB_VENDOR_MELCO,  USB_PRODUCT_MELCO_DUBPXXG, RID_WILDCARD,
          UMASS_PROTO_SCSI | UMASS_PROTO_BBB,
          FORCE_SHORT_INQUIRY | NO_START_STOP | IGNORE_RESIDUE
@@ -1084,7 +1106,8 @@
        /* Get the maximum LUN supported by the device.
         */
-       if ((sc->proto & UMASS_PROTO_WIRE) == UMASS_PROTO_BBB)
+       if ((sc->proto & UMASS_PROTO_WIRE) == UMASS_PROTO_BBB &&  /* XXX HACK */
+           !(sc->quirks & NO_GETMAXLUN))
                sc->maxlun = umass_bbb_get_max_lun(sc);
        else
                sc->maxlun = 0;


> I'll add it and close the PR.

You'll also need to add a couple vendor IDs to usbdevs, or the build
will fail again.  That's part of the giant diff I have, and as imp@
has been active on that front, I'm not sure how that is progressing.
My line to add the vendor ID is surrounded by some 50 lines worth of
diff, and the section of diff which adds the product is next to another
vendor/product new entry that could potentially break the build.

Also, the above cut-n-paste diff is part of a larger diff that makes
current's umass.c a plug-in replacement for 4.x

If a grep for `MAXTOR' in usbdevs returns the above added definitions,
you're in luck; else it's probably easiest to apply them by hand to
usbdevs from my latest diffs.


All this does is (for me) to speed up boot by a good deal of time.
Whether the Original Poster would still see the Original Problem, I
cannot say -- perhaps it's also an interaction with the particular
chipset in use.  You'll need to get a followup from the OP before
safely closing the PR.  (There have been enough changes since that
time that whatever it was could have been solved elsewhere, anyway)



Also, if you're wanting to update quirk entries in umass.c, I needed
to do the following to get one USB-IDE device to cooperate (in 4.x,
based on -current code, not tested on other than 4.x; not sure if the
original quirk entry was adequate for all devices, or if this particular
adapter is more broken than normal, or if -current doesn't have this
problem to start with)...

(tab damage)
@ -312,6 +321,9 @@
 #      define NO_INQUIRY               0x0400
        /* Device cannot handle INQUIRY EVPD, return CHECK CONDITION */
 #      define NO_INQUIRY_EVPD          0x0800
+/* XXX */
+       /* The device uses a weird CSW TAG. */
+#      define WRONG_CSWTAG             0x1000
 };

 Static struct umass_devdescr_t umass_devdescrs[] = {
@@ -419,7 +437,8 @@
        },
        { USB_VENDOR_SCANLOGIC, USB_PRODUCT_SCANLOGIC_SL11R, RID_WILDCARD,
          UMASS_PROTO_ATAPI | UMASS_PROTO_BBB,
-         NO_INQUIRY
+/* XXX */
+         NO_INQUIRY | WRONG_CSWTAG
        },
        { USB_VENDOR_SHUTTLE, USB_PRODUCT_SHUTTLE_EUSB, RID_WILDCARD,
          UMASS_PROTO_ATAPI | UMASS_PROTO_CBI_I,
@@ -1422,6 +1444,8 @@
 {
        struct umass_softc *sc = (struct umass_softc *) priv;
        usbd_xfer_handle next_xfer;
+/* XXX HACK */
+       int Residue;

        KASSERT(sc->proto & UMASS_PROTO_BBB,
                ("%s: umass_bbb_state: wrong sc->proto 0x%02x\n",
@@ -1595,7 +1619,12 @@
                    UGETDW(sc->csw.dCSWSignature) == CSWSIGNATURE_OLYMPUS_C1)
                        USETDW(sc->csw.dCSWSignature, CSWSIGNATURE);

-               int Residue;
+/* XXX HACK */
+               /* Translate invalid command-status tags */
+               if (sc->quirks & WRONG_CSWTAG)
+                       USETDW(sc->csw.dCSWTag, UGETDW(sc->cbw.dCBWTag));
+/* XXX */
+
                Residue = UGETDW(sc->csw.dCSWDataResidue);
                if (Residue == 0 &&
                    sc->transfer_datalen - sc->transfer_actlen != 0)

(relocation of Residue needed to quiet the 4.x compiler)
Stupid adapter won't see more than 131GB-ish worth of disk, however;
code for this quirk taken from NetBSD

Again, no guarantee that I haven't failed to include all relevant parts
of the diff needed to get a successful build.



> >(I don't use EHCI due to eventual problems with it after some time; I need
> >the stability, as of a few months ago)

> ehci is improving.. transacions can now time out without crashing the 
> system..

My bad ehci/umass experience was also with NetBSD, after most of
the commits from Charles Hannum that were supposed to fix the problem.
I'll try again when I don't risk losing several days of work that I've
been trying to complete uninterrupted for the last weeks without
success.  (Normal people would have thrown in the towel by now)


thansk
barry bouwsma