usb/101752: [panic] 6.1-RELEASE kernel panic on usb device insertion

[jes]

The following reply was made to PR usb/101752; it has been noted by GNATS.

From: jes <jes at jes-2.demon.nl>
To: bug-followup at freebsd.org, jes at jes-2.demon.nl
Cc: jes at jes-2.demon.nl
Subject: Re: usb/101752: [panic] 6.1-RELEASE kernel panic on usb device insertion
Date: Sun, 13 Aug 2006 15:51:26 +0200

 I built a test kernel with USB_DEBUG and got a bit more information prio=
 r  =
 
 to a panic:
 
 
 Aug 13 02:14:11 jes-2 kernel: usbd_new_device bus=3D0xc3741000 port=3D1 =
  =
 
 depth=3D1 speed=3D2
 
 Aug 13 02:14:11 jes-2 kernel: usbd_new_device: adding unit addr=3D2,  =
 
 rev=3D200, class=3D0, subclass=3D0, protocol=3D0, maxpacket=3D64, len=3D=
 18, speed=3D2
 
 Aug 13 02:14:11 jes-2 kernel: usbd_new_device: new dev (addr 2),  =
 
 dev=3D0xc4093380, parent=3D0xc373f580
 
 Aug 13 02:14:11 jes-2 kernel: usbd_probe_and_attach: trying device  =
 
 specific drivers
 
 Aug 13 02:14:11 jes-2 kernel: usbd_probe_and_attach: no device specific =
  =
 
 driver found
 
 Aug 13 02:14:11 jes-2 kernel: usbd_probe_and_attach: looping over 2  =
 
 configurations
 
 Aug 13 02:14:11 jes-2 kernel: usbd_set_config_index: status=3D0x0001,  =
 
 error=3DNORMAL_COMPLETION
 
 Aug 13 02:14:11 jes-2 kernel: usbd_set_config_index: (addr 1) cno=3D2  =
 
 attr=3D0xc0, selfpowered=3D1, power=3D100
 
 Aug 13 02:14:11 jes-2 kernel: usbd_set_config_index: set config 1
 
 Aug 13 02:14:12 jes-2 kernel: umass1: CASIO COMPUTER CASIO QV DIGITAL, r=
 ev  =
 
 2.00/10.00, addr 2
 
 Aug 13 02:14:12 jes-2 kernel: umass1: 8070i (ATAPI) over Bulk-Only; quir=
 ks  =
 
 =3D 0x0000
 
 Aug 13 02:14:12 jes-2 kernel: umass1:4:1:-1: Attached to scbus4
 [panic]
 Aug 13 02:16:46 jes-2 syslogd: restart
 
 I tried adding CAM_DEBUG, but things went a bit awry. If you set any of =
 the
 CAM_DEBUG_FLAGS, you get a kernel panic (RELENG 6.1_1) at startup time, =
  =
 
 before even the hard discs are enabled.
 
 options         CAMDEBUG
 
 options         CAM_DEBUG_BUS=3D1
 
 options         CAM_DEBUG_TARGET=3D-1
 
 options     CAM_DEBUG_LUN=3D-1
 
 options         CAM_DEBUG_FLAGS=3D56
 
 # CAM_DEBUG_INFO|CAM_DEBUG_TRACE|CAM_DEBUG_SUBTRACE|CAM_DEBUG_CDB
 
 The panic is that the GIANT lock is not held by the code, at cam_xpt.c:4=
 193
 
 No discs mean no crash dump.
 
 Setting CAM_DEBUG_FLAGS to 0 avoids a panic, but is hardly informative.
 I started doing a binary search to see if I could find flags which don't=
   =
 
 panic the kernel adn still produce some output, but the first two  =
 
 attempts, with flags 7 and flags 56 bot paniced. I assume that doing  =
 
 console output releases the GIANT lock
 
 
 -- =
 
 Jim Segrave  - jes at jes-2.demon.nl