It looks too me that this would only be a problem if the master has it's
debug level set to 0 and the clients have it set to something else.
Its still a serious bug though, but it explains why I didn't see it
before.
--
Hugh Greenberg <[EMAIL PROTECTED]>
On Thu, 2008-11-06 at 12:17 -0800, ron minnich wrote:
> This is when we're still trying to open/write the log!
>
> on client:
> write(3, "\'\0\0\0v\0\0\1\0\0\0\0\0\0\0\0\0\0\0\20\0\0\0listen on
> 54183\n", 39)
> poll([{fd=3, events=POLLIN|POLLOUT, revents=POLLOUT}, {fd=5,
> events=POLLIN|POLLO
> poll([{fd=3, events=POLLIN, revents=POLLIN}, {fd=5, events=POLLIN|
> POLLOUT}], 2,
> read(3, "Client 10.1.1.73!43082: listen on 54183\n", 32792) = 40
> shutdown(3, 2 /* send and receive */) = 0
> close(3) = 0
> write(2, "Error: invalid fcall size greater than msize: 5\n", 48) = 48
> stat64(".", {st_mode=S_IFDIR|0755, st_size=0, ...}) = 0
> write(2, "Fatal error: Could not find file:
> vnfs/caos-nsa-node-1.0rc1-1.stateles
> exit_group(-1) = ?
>
> now we're tracing server:
> poll([{fd=4, events=POLLIN|POLLOUT}, {fd=0, events=POLLIN}, {fd=1,
> events=POLLIN
> }, {fd=5, events=POLLIN}, {fd=6, events=POLLIN}, {fd=7,
> events=POLLIN}, {fd=8, e
> vents=POLLIN}, {fd=9, events=POLLIN}, {fd=2, events=POLLIN,
> revents=POLLIN}], 9,
> 300000) = 1
> read(2, "\'\0\0\0v\0\0\1\0\0\0\0\0\0\0\0\0\0\0\20\0\0\0listen on 50621
> \n", 32792
> ) = 39
> write(2, "Client 10.1.1.73!41808: listen on 50621\n", 40) = 40
> time(NULL) = 1225977306
> write(2, "\v\0\0\0w\0\0\20\0\0\0", 11) = 11
> poll([{fd=4, events=POLLIN|POLLOUT}, {fd=0, events=POLLIN}, {fd=1,
> events=POLLIN
> }, {fd=5, events=POLLIN}, {fd=6, events=POLLIN}, {fd=7,
> events=POLLIN}, {fd=8, e
> vents=POLLIN}, {fd=9, events=POLLIN}, {fd=2, events=POLLIN|POLLOUT,
> revents=POLL
> IN|POLLOUT}], 9, 300000) = 1
> read(2, "", 32792) = 0
> close(2) = 0
>
> note that we are doing I/O to the client on fd 2. Note also that we
> appear to be dumping log data of some sort on fd 2. This log data is
> clearly not 9p. I assume there is an fprintf(2 in there somewhere.
>
> This trashes the connection with the client.
>
