On Tue, 19 Dec 2006 12:12:59 -0800, Edward Jaffe <[EMAIL PROTECTED]> wrote:
>... >> To disable Segmentation Offload, it is >> recommended that the following maintenance be applied >> and then code NOSEGMENTATIONOFFLoad on the >> GLOBALCONFIG statement in the TCPIP PROFILE. >> > >This issue has been around foe a while. Originally, the segmentation >offload could be disabled only through an operator command. Later, an >APAR was taken to make it configurable via the GLOBALCONFIG statement. >The PSP bucket gets updated from time-to-time, but IBM still recommends >the feature be disabled. >... I haven't checked into this problem for a few months but both the bug and the circumvention are a bit problematic. There were 2 fixes available to turn off the OFFLOAD function. The first fix implemented support for the "F tcp_proc,NOSEGOFFLOAD" command. It executes (success or failure) silently - no message. The 2nd fix implemented the GLOBALCONFIG support and removed the command support. Prior to that 2nd fix the command succeeds silently; afterwords it fails silently. You've gotta know your maintenance level. (We don't have that 2nd fix on yet. It may not be quite as stealthful as I've heard.) For a while IBM was requesting customers to test the microcode fix on unimportant (i.e., test) OSA ports. That was when they thought the fix was going to be available soon. And that assumed that customers had OSA ports that were not shared beween test and production LPARs. (That may be true in some shops. Not ours.) I haven't heard such requests lately. Pat O'Keefe ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
