Hi all, I sent this message a while ago, but I only saw it appear on opensolaris-discuss and not here on driver-discuss. It got no responses, so now that I have more info, I thought I'd try again.
I've since installed several versions of RHEL (4u7, 5u2, etc.) on these same blades and bladecenters, and the SOL console works fine all through boot. I haven't gone back to try S10, but I've continued to test sNV (SXCE) and shortly after the kernel banner, while configuring /dev the SOL console times out, gets disconnected, and can't be reconnected until the blade is reset. I also want to go back and test this with versions of sNV before the /dev/bmc driver was integrated (or was it updated?) recently. ANyone know what build that might have been? Or, were there any changes to the bge network driver that might have changed how it is reset during this portion of booting? -Kyle Forwarded Message: > Hi all, > > I have an IBM BladeCenter with several HS20 blades. Each of these has > a BMC module and one of the ways to use the console of the blade is to > redirect the console to the blade's serial port which the BMC module > then translates to SOL protocol, which you can access through the > network. > > I am having a problem (on sNV97 and sNV98) maintaining the console > access this way when running Solaris on these blades (I am going to > test linux next.) The SOL console works great through POST, all the > BIOS menus, grub, and even the initial Kernel startup. But as soon as > the Solaris kernel boots (banner appears) and the message about > configuring /dev comes up, something in Solaris causes the BMC to drop > the SOL connection and refuses to restart it (reports 'SOL not > ready'.) As soon as the blade is rebooted and POST starts it starts > working again. > > > There is one known similar problem. Since the network connection to > the BMC is shared with the first ethernet port on the blade, it is > known that PXE/DHCP booting the blade using the first ethernet > interface, resets the NIC and will cause the active SOL console > session to be disconnected just like what I'm describing above. > > This isn't my problem (I'm not PXE/DHCP booting, and when I do I use > the second ethernet not the first,) but it sounds very similar. So I'm > wondering 2 things: > > 1.) Could the Broadcom ethernet driver be 'reseting' the first NIC > exactly the same way the PXE/DHCP boot would, and therefore be > triggering the same issue? > > 2.) I seem to recall a recent integration or change involving a > Solaris driver to interact with the BMC module. Could that or > something else in the kernel be resetting the BMC directly, or > otherwise causing the SOL connection to drop? > > I need to go back and test earlier sNV builds and s10 (which IBM > claims works fine) but I thought I'd ask here first. > > -Kyle _______________________________________________ driver-discuss mailing list [email protected] http://mail.opensolaris.org/mailman/listinfo/driver-discuss
