I seem to have run into a problem reading data from a post operation. The data being posted is an XMLRPC request and the sizes range from about 100 to 1000 bytes - i.e. not that much data, really.
Using the ns_conncptofp to copy the content to a temporary file I routinely get the error "resource temporarly unavailable". Which we all know is typically cause by an i/o operation on a non-blocking descriptor when, for some reason, reading or writing cannot be done at that time. I've added a bit of debugging code to track down why this is happining and here is what I get from my log files. [31/Oct/2001:17:34:01][11632.16][-conn1-] Notice: Length 547 [31/Oct/2001:17:34:01][11632.16][-conn1-] Notice: Allocating space for 547 bytes [31/Oct/2001:17:34:01][11632.16][-conn1-] Notice: reading 547 bytes from socket [31/Oct/2001:17:34:01][11632.16][-conn1-] Notice: Ns_SockRecv: reading 16000, buffer 2747b4, timeout 30 [31/Oct/2001:17:34:01][11632.16][-conn1-] Notice: Ns_SockRecv: read -1, errno 11 [31/Oct/2001:17:34:01][11632.16][-conn1-] Notice: Ns_SockWait: 30 [31/Oct/2001:17:34:01][11632.16][-conn1-] Notice: Ns_SockWait: seting up fdset [31/Oct/2001:17:34:01][11632.16][-conn1-] Notice: Ns_SockWait: selecting for read [31/Oct/2001:17:34:01][11632.16][-conn1-] Notice: Ns_SockWait: read select done, result 1 [31/Oct/2001:17:34:01][11632.16][-conn1-] Notice: Ns_SockWait: select returned 1, errno 0 [31/Oct/2001:17:34:01][11632.16][-conn1-] Notice: Ns_SockWait: returning OK [31/Oct/2001:17:34:01][11632.16][-conn1-] Notice: Ns_SockRecv: Rereading socked 7, buffer 2747b4, toread 16000 [31/Oct/2001:17:34:01][11632.16][-conn1-] Notice: Ns_SockRecv: read -1, errno 11 [31/Oct/2001:17:34:01][11632.16][-conn1-] Notice: Ns_SockRecv: returning -1 [31/Oct/2001:17:34:01][11632.16][-conn1-] Notice: -1 bytes, Errno: 11 [31/Oct/2001:17:34:01][11632.16][-conn1-] Notice: Results were -1 bytes read (-1 means error) [31/Oct/2001:17:34:01][11632.16][-conn1-] Notice: Results set to error message The lines that have Ns_SockRecv or Ns_SockWait were printed from within those functions inside of nsd/sock.c It looks like the first recv in Ns_SockRecv is returning EWOULDBLOCK, so we decide to call Ns_SockWait which performs a select on the socket. The select returns without error, indicating that one file descriptor is ready to be read from. Ns_SockWait then returns an OK since select said there is data ready to be read on the socket. However, when we then attempt the second recv in Ns_SockRecv we again get an error EWOULDBLOCK (which is the same as EAGAIN which is ernno = 11). What's happining here? Has anyone seen this before? Here is some info on the system we are running: ~> uname -a SunOS hostname 5.8 Generic_108528-07 sun4u sparc SUNW,UltraAX-i2 ~> gcc -v Reading specs from /usr/local/lib/gcc-lib/sparc-sun-solaris2.8/2.95.3/specs gcc version 2.95.3 20010315 (release) Truthfully, this is looking like something more like a kernel or gcc bug to me - the aolserver code seems to be ok, but I want to check here to see if anyone else has seen this behavior. Thanks, Matthew
