Re: cvs commit: src/sys/bus/usb usb_ethersubr.c src/sys/kern kern_poll.c uipc_msg.c src/sys/net netisr.c netisr.h src/sys/net/ppp if_ppp.c src/sys/netgraph/netgraph ng_base.c src/sys/netinet if_ether.c ip_demux.c ip_input.c tcp_subr.c ...

[YONETANI Tomokazu <qhwt+dragonfly-commits@xxxxxxxxxx>]

Hi.

On Fri, Apr 09, 2004 at 03:34:10PM -0700, Jeffrey Hsu wrote:
> hsu         2004/04/09 15:34:10 PDT
> 
> DragonFly src repository
> 
>   Modified files:
>     sys/bus/usb          usb_ethersubr.c 
>     sys/kern             kern_poll.c uipc_msg.c 
>     sys/net              netisr.c netisr.h 
>     sys/net/ppp          if_ppp.c 
>     sys/netgraph/netgraph ng_base.c 
>     sys/netinet          if_ether.c ip_demux.c ip_input.c 
>                          tcp_subr.c 
>     sys/netinet6         ip6_input.c 
>     sys/netproto/atalk   aarp.c ddp_input.c 
>     sys/netproto/atm     atm_subr.c 
>     sys/netproto/ipx     ipx_input.c 
>     sys/netproto/natm    natm.c 
>     sys/netproto/ns      ns_input.c 
>   Log:
>   Push the lwkt_replymsg() up one level from netisr_service_loop() to
>   the message handler so we can explicitly reply or not reply as appropriate.
>   
>   Revision  Changes    Path
>   1.8       +4 -0      src/sys/bus/usb/usb_ethersubr.c
>   1.9       +4 -2      src/sys/kern/kern_poll.c
>   1.6       +5 -6      src/sys/kern/uipc_msg.c
>   1.11      +2 -26     src/sys/net/netisr.c
>   1.12      +2 -2      src/sys/net/netisr.h
>   1.15      +4 -3      src/sys/net/ppp/if_ppp.c
>   1.11      +3 -1      src/sys/netgraph/netgraph/ng_base.c
>   1.11      +4 -0      src/sys/netinet/if_ether.c
>   1.16      +24 -0     src/sys/netinet/ip_demux.c
>   1.17      +7 -5      src/sys/netinet/ip_input.c
>   1.21      +2 -3      src/sys/netinet/tcp_subr.c
>   1.13      +21 -21    src/sys/netinet6/ip6_input.c
>   1.8       +3 -1      src/sys/netproto/atalk/aarp.c
>   1.7       +4 -2      src/sys/netproto/atalk/ddp_input.c
>   1.11      +1 -0      src/sys/netproto/atm/atm_subr.c
>   1.9       +8 -6      src/sys/netproto/ipx/ipx_input.c
>   1.15      +7 -2      src/sys/netproto/natm/natm.c
>   1.10      +9 -8      src/sys/netproto/ns/ns_input.c

After this commit(or maybe the next one saying ``Send connects
to the right processor''), mpd stopped working. Although
it doesn't panic, mpd gets stuck and become non-killable
process; ddb showed the following trace:

lwkt_switch(c5d75bc0)
lwkt_default_waitport(c5d75bd0, c8a56ac8, c7ec1ce0, c8a56b20, c01764a0)
lwkt_domsg(c021e4f0, c8a56ac8, 0, 0, c7ec1ce0)
so_pru_send(c7ec1ce0, 0, c5dbb100, c0f81a90, 0)
sosend
kern_sendmsg
sendto
syscall2

This is with the latest source(at least with after the removal of
TCP_DISTRIBUTED_TCBINFO). Does the recent changes require recompilation
of user-land applications?

By the way, is there a way to specify from which process to show the
trace in gdb -k? For example, mpd had pid 36 in the above trace, but
if I manually panic, savecore the dump and feed it to gdb -k, I don't
know how to specify the pid in gdb. It's too late to simply set a
breakpoint on lwkt_default_waitport after mpd got stuck because, but
if I set the breakpoint BEFORE mpd hangs, maybe I have to skip too
many false-positives before reaching the target.