Dave Miner wrote: > Ethan Quach wrote: >> >> Dave Miner wrote: >>> Caimaniacs, >>> >>> In looking at some solutions for 8558 which would allow SUNWemlxs to >>> stay on the media, I'm seeing frequent hangs in the disks screen as >>> it's waiting for target discovery to complete. As I've been looking >>> through some code to figure out how it could hang in order to >>> understand what I might do about it, I notice that libdiskmgt has an >>> environment variable, _LIBDISKMGT_INSTALL, which can be used to >>> prevent libdiskmgt from hooking into the sysevents for notification >>> of new devices. Digging around, I see that the old Solaris installer >>> does use this variable, but we're not. Im assuming in the Solaris >>> case it was to avoid putting something additional into the miniroot, >>> but I haven't gone off and researched that. Anyone have a reason why >>> we aren't using it in Caiman (and why we shouldn't, since it seems >>> to reduce memory usage of the installer somewhat)? I didn't find >>> anything in the target discovery design doc. >> >> Didn't we have a discussion about this way back when, and decided that >> if a user were to plug in a USB drive for example, they should be >> able to >> go back a screen, then go forward, and it should show up as a device >> to install on in the disk screen? (But I'm not even sure if libdiskmgmt >> would need sysevents to get notified of such things...) >> > > We did discuss it at one time. I seem to recall that we decided not > to do that because we couldn't hook it up as an event-driven mechanism > all the way up, and it could take a significant amount of time to > rediscover, which would make the GUI pretty frustrating in some cases. > I think we also considered, and rejected, a "reload" button on that > screen and decided to just leave it as restarting the installer to get > a rediscovery to happen. Anyway, that's how it works according to my > testing with 111a. So it doesn't seem the sysevent stuff is doing us > any good. I'll wait for Sarah to speak up since she's the libdiskmgt > expert. Hi Dave,
Sorry for the delay...I know that a lot of changes have been made in the HAL layer for removable media management. So, even if we did have the discussion to allow sysevent capability in libdiskmgt, to be honest, it isn't likely to work very well, and may be causing the hang you are experiencing. Garrett's recent bug had to do with libdiskmgt trying to discover his USB device. Not a sysevent issue, but just to point out the removable media issues. In the old installer, the dependency on libdiskmgt was introduced as as result of using 'newfs'. We didn't do device in use checking directly. The flag, _LIBDISKMGT_INSTALL, was introduced to ensure sysevents didn't start during the install. I don't think the sysevent stuff is reliable enough at this point, in particular for USB devices, that we should leave it enabled in libdiskmgt. sarah *** > > Dave > > _______________________________________________ > caiman-discuss mailing list > caiman-discuss at opensolaris.org > http://mail.opensolaris.org/mailman/listinfo/caiman-discuss
