[PATCH] Fix unloading modules in driver-updates (#1085099)

Brian C. Lane bcl at redhat.com
Mon Apr 21 18:42:13 UTC 2014


On Mon, Apr 21, 2014 at 10:31:31AM -0400, Will Woods wrote:
> An unintended consequence of commit de2d663 is that driver-updates won't
> unload *any* modules at all: /tmp/dd_modules (the list of "early"
> modules that shouldn't be removed) is now created *after* we load all
> modules.)
> 
> This patch makes driver-updates attempt to remove any driver that's in
> the update disk(s) it loads.
> 
> First we make dd_extract() return a list of the firmware/module files
> that were in the driver disk it extracted.
> 
> Next process_dd() builds a list of those for each extracted driver and
> passes that to reload_modules().
> 
> reload_modules() maps those filenames to module names, and adds them to
> the list of modules to be unloaded.
> 
> NOTE: Probably the correct behavior (and the way it worked in RHEL6)
> would be to revert the change to /tmp/dd_modules and fix that bug in a
> different way (e.g. by tracking driver disks by UUID instead of device
> name, so they don't get loaded twice if the device name changes).
> 
> That requires a bunch of refactoring work; this is a stopgap solution.

Ack. Thanks for fixing this.

-- 
Brian C. Lane | Anaconda Team | IRC: bcl #anaconda | Port Orchard, WA (PST8PDT)


More information about the anaconda-patches mailing list