EHCI problem on -current, and unknown devs

[Stephane E. Potvin]

John Baldwin wrote:
> On Thursday 30 November 2006 22:33, Stephane E. Potvin wrote:
>> John Baldwin wrote:
>>> On Wednesday 29 November 2006 17:39, Stephane E. Potvin wrote:
>>>> John Baldwin wrote:
>>>>> On Thursday 16 November 2006 03:34, Stephane E. Potvin wrote:
>>>>>> Doug Barton wrote:
>>>>>>> [ No response from the -usb list, so I'll try here. ]
>>>>>>>
>>>>>>> Howdy,
>>>>>>>
>>>>>>> I have a new Dell Latitude D620, and using recent -current I get this:
>>>>>>>
>>>>>>> ehci0: <Intel 82801GB/R (ICH7) USB 2.0 controller> mem
>>>>>>> 0xffa80000-0xffa803ff irq 20 at device 29.7 on pci0
>>>>>>> ehci0: Could not map memory
>>>>>>> device_attach: ehci0 attach returned 6
>>>>>>>
>>>>>>> FWIW, I also saw a user on the -stable list report this same problem.
>>>>>>> I'm assuming this is bad, but what to do about it?
>>>>>>>
>>>>>>>
>>>>>>> I'm also getting some unknown devices:
>>>>>>>
>>>>>>> uhub4: <vendor 0x413c product 0xa005, class 9/0, rev 2.00/50.18, addr
>>>>>>> 2> on uhub0
>>>>>>> uhub5: <vendor 0x0b97 product 0x7761, class 9/0, rev 1.10/1.10, addr
>>>>>>> 3> on uhub4
>>>>>>> ugen2: <vendor 0x413c product 0x8103, class 224/1, rev 2.00/24.22,
>>>>>>> addr 6> on uhub4
>>>>>>> uhub6: <vendor 0x413c product 0x0058, class 9/0, rev 2.00/0.00, addr
>>>>>>> 2> on uhub3
>>>>>>>
>>>>>>> Vendor ID 0x413c is Dell, and is already in the usbdevs file. The
>>>>>>> other vendor is O2, and I got their ID for the attached patch from the
>>>>>>> list at http://www.usb.org/developers/tools. The 0x0058 device is
>>>>>>> already in our usbdevs file, it's the port replicator (docking
>>>>>>> station) that the laptop is plugged into currently. I added the 0x8103
>>>>>>> device in the attached patch based on an entry in the NetBSD usbdevs
>>>>>>> file. (The entry makes sense to me as well, since I have one of 
> those.)
>>>>>>> So I have two questions ... for the devices (and vendors) that are
>>>>>>> already in my local usbdevs file, why are they still showing up by ID?
>>>>>>> And how do I find the device IDs for the two unknown devices? I
>>>>>>> imagine that the O2 device is related to my built in smart card
>>>>>>> reader, not sure about the other one.
>>>>>>>
>>>>>> Hi Doug,
>>>>>>
>>>>>> I had the same problem with my Dell Inspiron 9400 and fixed it using 
> the 
>>>>>>   following patch:
>>>>>>
>>>>>> Index: sys/kern/subr_rman.c
>>>>>> ===================================================================
>>>>>> RCS file: /home/FreeBSD/ncvs/src/sys/kern/subr_rman.c,v
>>>>>> retrieving revision 1.53
>>>>>> diff -u -r1.53 subr_rman.c
>>>>>> --- sys/kern/subr_rman.c        11 Sep 2006 19:31:52 -0000      1.53
>>>>>> +++ sys/kern/subr_rman.c        2 Nov 2006 03:05:34 -0000
>>>>>> @@ -170,7 +170,7 @@
>>>>>>
>>>>>>          /* Skip entries before us. */
>>>>>>          for (s = TAILQ_FIRST(&rm->rm_list);
>>>>>> -            s && s->r_end + 1 < r->r_start;
>>>>>> +            s && s->r_end < r->r_start;
>>>>>>               s = TAILQ_NEXT(s, r_link))
>>>>>>                  ;
>>>>>>
>>>>>> This code was added in revision 1.53 when support for intelligent 
>>>>>> merging was added. When r_end is equal to UINT_MAX, adding one will 
>>>>>> cause it to overflow, creating quite a mess in the entries ordering. 
> I'm 
>>>>>> not sure the fix I did is completely correct as I didn't had time to 
>>>>>> check if the +1 is needed at all in this case. At least it fixed the 
>>>>>> problem for me.
>>>>> It is needed to avoid overlaps.  Can you show me which 
>>> rman_manage_region() is 
>>>>> breaking?  Rather, enable the DPRINTF() in rman_manage_region() and 
>>> provide 
>>>>> the output?
>>>>>
>>>> Here's the output of booting an unmodified kernel with debug.rman_debug=1
>>> Try this:
>>>
>> I had to modify the patch a little to make it compile:
>>
>> When I try to boot after patching nearly all the drivers fail to 
>> allocate their resources, making the laptop unable to mount root as the 
>> ata driver failed to attach. I can't provide any log yet as I don't have 
>> a serial port on the laptop to grab the output.
> 
> Sign bug on my part.  try this:
> 
[... patch removed]

I had to modify the patch a little again to make it compile. With this 
slight modification (changing rm->link to rm->list), the patch seems to 
be working fine and I got back my ehci controller.

Index: subr_rman.c
===================================================================
RCS file: /home/FreeBSD/ncvs/src/sys/kern/subr_rman.c,v
retrieving revision 1.53
diff -u -r1.53 subr_rman.c
--- subr_rman.c 11 Sep 2006 19:31:52 -0000      1.53
+++ subr_rman.c 1 Dec 2006 15:25:59 -0000
@@ -61,6 +61,7 @@
  #include <sys/param.h>
  #include <sys/systm.h>
  #include <sys/kernel.h>
+#include <sys/limits.h>
  #include <sys/lock.h>
  #include <sys/malloc.h>
  #include <sys/mutex.h>
@@ -169,10 +170,12 @@
         mtx_lock(rm->rm_mtx);

         /* Skip entries before us. */
-       for (s = TAILQ_FIRST(&rm->rm_list);
-            s && s->r_end + 1 < r->r_start;
-            s = TAILQ_NEXT(s, r_link))
-               ;
+       TAILQ_FOREACH(s, &rm->rm_list, r_link) {
+               if (s->r_end == ULONG_MAX)
+                       break;
+               if (s->r_end + 1 >= r->r_start)
+                       break;
+       }

         /* If we ran off the end of the list, insert at the tail. */
         if (s == NULL) {

Thanks a lot!

Steph