Re: cvs commit: src/sys/kern uipc_socket.c src/sys/sys poll.h src/sys/vfs/fifofs fifo_vnops.c

[Matthew Dillon <dillon@xxxxxxxxxxxxxxxxxxxx>]

:Now I'm not longer able to boot off my 3ware controller.
:It starts to boot but goes into an infinite loop resetting itself....
:
:-DR

    Is it due to this commit or an earlier one?  e.g. if the 3ware controller
    needs ISA DMA space it could be the slab allocator changes that are 
    messing it up.

    I think you may have to revert your tree incrementally to find out 
    exactly where we broke the controller.  But before you do try going
    back to the original allocator and see if that fixes it (i.e. comment
    out the NO_KMEM_MAP and USE_SLAB_ALLOCATOR kernel options if you had set
    them).  If it is the ISA DMA space I will take a timeout from the VFS
    code and try to fix that more quickly then my current plan.

					-Matt
					Matthew Dillon 
					<dillon@xxxxxxxxxxxxx>


:On Wednesday, September 3, 2003, at 08:58 AM, Hiten Pandya wrote:
:
:> hmp         2003/09/03 05:58:05 PDT
:>
:>   Modified files:
:>     sys/kern             uipc_socket.c
:>     sys/sys              poll.h
:>     sys/vfs/fifofs       fifo_vnops.c
:>   Log:
:>   Introduce a new poll operation bit, `POLLINGIGNEOF'.  It is used for
:>   ignoring EOFs on a FIFO.  POLLIN/POLLRDNORM is converted withing the
:>   FIFO implementation to effect the correct behaviour.
:>
:>   Yes... POLLINIGNEOF flag name is a bit hard on the eyes. :-(
:>
:>   Obtained from:  	FreeBSD
:>
:>   More information:	FreeBSD GNATS (PR #34020)
:>
:>   Revision  Changes    Path
:>   1.10      +8 -1      src/sys/kern/uipc_socket.c
:>   1.4       +5 -0      src/sys/sys/poll.h
:>   1.9       +27 -5     src/sys/vfs/fifofs/fifo_vnops.c
:>
: