Attached is the reworked patch to take care of Andrew's suggestions -

1) Allocate the work struct dynamically in struct ti_ohci during  device 
probe, free it during device remove
2) In ohci1394_pci_remove, ensure queued work, if any, is flushed before 
the device is removed (As I understand, this should work for both device 
removal cases as well as module removal - correct?)
3) Rework the free_dma_rcv_ctx  - It is called in multiple contexts by 
different callers - teach it to free the dma according to caller's wish 
- either direct free where caller determines the context is safe to 
sleep OR delayed via schedule_work when caller wants to call it from a 
context where it cannot sleep.  So it now takes 2 extra arguments - 
method to free - direct or delayed and if it is to be  freed delayed, an 
appropriate work_struct.

Andrew - flush_scheduled_work does not touch work which isn't shceduled 
- so I don't think INIT_WORK in setup is necessary. Let me know if I am 
missing something here. (I tried INIT_WORK in ochi1394_pci_probe and 
putting flush_scheduled_work in ohci1394_pci_remove - It did not result 
in calling the work function after I did rmmod, since it wasn't scheduled.)

Parag