trouble sending very large rpc replies

[James Peach]

Hi guys,

I'm still slowly plugging away at integrating PCP support for
performance metrics. When the necessary management folk get back from
holidays in a couple of weeks, I'll be able to post some patches and
test programs.

In the meantime, I've hit what appears to be a snag in the RPC layer.
When the client fetches the counter names and help text registry
entries, these get quite large due to the number of counters we have.
The help text is ~100KiB on Linux and ~500KiB on IRIX.

I'm currently testing on Linux and it seems that send_trans_reply thinks
this is too large a chunk of data to send and I get the following
failure:
        
        [2005/12/24 13:26:42, 5] smbd/ipc.c:send_trans_reply(89)
        send_trans_reply: buffer 1024 too large
        [2005/12/24 13:26:42, 3] smbd/error.c:error_packet(146)
        error packet at smbd/ipc.c(97) cmd=37 (SMBtrans) STATUS_BUFFER_OVERFLOW
        
Is this behaviour to be expected? Is there something I can do to work
around it?

The test program in question does a RegQueryValueEx starting with a
16KiB buffer. It grows the buffer by 16KiB each time is gets a
ERROR_MORE_DATA return value.

I have a debug (5) log and a packet trace here:
	samba.org:~jpeach/logs/perfdump.{cap,log}.gz

I appreciate any comment/guidance/advice ...

-- 
James Peach | jpeach at samba.org