Tom Matthews wrote:
> Steffen,
>
>> I thought an unprivileged application should be able to set it to
>> [udp|tcp]_max_buf
>
> The application can request an SO_RECVBUF of anything up to the receive
> buffers highwater mark (tcp_recv_hiwat or udp_recv_hiwat).
>
> To go back to my original question, can I tune the hiwat value for a system
> using a mix of 1Gbit and 100Mbit interfaces (both with 2ms latency) to the
> region of 240KB, or will that somehow negatively impact the 100Mbit
> interfaces performance.
>
> Best regards, Tom.
>
>
> This message posted from opensolaris.org
> _______________________________________________
> networking-discuss mailing list
> [email protected]
On the Sun Ray server I am working on (I don't have priviliged access),
running S10, not OpenSolaris, I launched a ttcp receiver...
[EMAIL PROTECTED]: 405% pgrep ttcp
42838
[EMAIL PROTECTED]: 406% pfiles 42838
42838: ttcp -r -p 1111 -b 70000
Current rlimit: 128 file descriptors
0: S_IFCHR mode:0620 dev:350,0 ino:12582922 uid:10789 gid:7 rdev:24,3
O_RDWR
/devices/pseudo/[EMAIL PROTECTED]:3
1: S_IFCHR mode:0620 dev:350,0 ino:12582922 uid:10789 gid:7 rdev:24,3
O_RDWR
/devices/pseudo/[EMAIL PROTECTED]:3
2: S_IFCHR mode:0620 dev:350,0 ino:12582922 uid:10789 gid:7 rdev:24,3
O_RDWR
/devices/pseudo/[EMAIL PROTECTED]:3
3: S_IFSOCK mode:0666 dev:356,0 ino:25742 uid:0 gid:0 size:0
O_RDWR
SOCK_STREAM
SO_SNDBUF(49152),SO_RCVBUF(70216),IP_NEXTHOP(0.1.18.72)
sockname: AF_INET 0.0.0.0 port: 1111
[EMAIL PROTECTED]: 407% ndd -get /dev/tcp tcp_recv_hiwat
49152
[EMAIL PROTECTED]: 408% ndd -get /dev/tcp tcp_max_buf
1048576
[EMAIL PROTECTED]: 418% ttcp -r -p 1111 -b 1048576
ttcp-r[1]: buflen=8192, nbuf=2048, align=16384/0, port=1111,
sockbufsize=1048576 tcp
ttcp-r[1]: socket
ttcp-r[1]: rcvbuf
[EMAIL PROTECTED]: 419%
[EMAIL PROTECTED]: 419% ttcp -r -p 1111 -b 1048577
ttcp-r[1]: buflen=8192, nbuf=2048, align=16384/0, port=1111,
sockbufsize=1048577 tcp
ttcp-r[1]: socket
ttcp-r[1]: setsockopt: rcvbuf: No buffer space available
errno=132
[EMAIL PROTECTED]: 420%
_______________________________________________
networking-discuss mailing list
[email protected]
