Red Hat Bugzilla – Bug 464750
RFE: On kernel install /boot status should be checked
Last modified: 2009-11-27 07:41:51 EST
Description of problem:
Well, this is more an RFE than a problem, albeit sometimes it is a problem...
I've the /boot partition, for several reasons, set to "ro" (someone recommends even not to auto-mount it).
When a kernel install/update/remove occurs, sometimes I forgot to set the /boot to "rw" before starting the operations, with obvious results...
It would be nice, if possible, to check /boot as first operation during kernel install/update/remove and, eventually, bail out with an error, before doing anything else.
Thanks a lot in advance.
This should be done by the package manager if at all.
(In reply to comment #1)
> This should be done by the package manager if at all.
I think you might be right.
One other potential use case is to have /usr mounted "ro".
Fixed in rpm-4.7.2 in rawhide now, install attempt on read-only filesystems now errors out early as it should.
thanks for the update.
I've got anyway one question.
What happens, like in case of kernel install/update, where something goes in /boot, something else in /lib/modules and other in /usr, if the different folders belong to different mount points with different rw/ro permissions?
My case would be (for the kernel):
/usr ro (formerly rw)
Does "rpm" check all the folders _before_ installing anything?
Or succeeding in /lib/modules will make it assume the rest is OK? In the hypothesis /lib/modules is the first folder used.
Obviously all filesystems are checked and must have enough room for install to even start. This has always been the case, but until now read-only filesystems were not handled correctly.