Most recent kernel where this bug did not occur: unknown

(kernel used was actually 2.6.17.11 plus 1394 driver updates from end of August
2006)

Problem Description: Some of libraw1394's ARM related requests cause raw1394 to
call sleeping functions while interrupts are disabled. This leads to warnings (I
suppose if the kernel was compiled with sleep-inside-spinlock checking) or to
lockups if the sleeping functions actually need to sleep.

Sep  7 20:47:50 shuttle kernel: BUG: sleeping function called from invalid
context at arch/i386/lib/usercopy.c:607
Sep  7 20:47:50 shuttle kernel: in_atomic():1, irqs_disabled():1
Sep  7 20:47:50 shuttle kernel:  <c011a4b1> __might_sleep+0xa1/0xc0  <c0158967>
__vmalloc+0x27/0x30
Sep  7 20:47:50 shuttle kernel:  <c01f5371> copy_to_user+0x21/0x80  <f89a58f5>
arm_register+0x405/0x4c0 [raw1394]
Sep  7 20:47:50 shuttle kernel:  <f89a686e> raw1394_write+0xae/0xe0 [raw1394] 
<c0165fc0> vfs_write+0xd0/0x1e0
Sep  7 20:47:50 shuttle kernel:  <c01661a1> sys_write+0x51/0x80  <c010315b>
sysenter_past_esp+0x54/0x75

Steps to reproduce: Run any libraw1394 client with ARM handling. The above
output occured at startup of Oracle's Endpoint.

Actual lockups were reported too:
http://article.gmane.org/gmane.linux.kernel.firewire.user/1961