| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
| |
Call the umount(8) command when unmounting filesystems through isys so
that /etc/mtab is updated.
|
| |
|
|
|
|
| |
Turn doPwMount() in to a mount command wrapper. We're already calling
mount(8) and capturing error messages so we can pass them back up to our
Python code, so the same for unmounting filesystems.
|
| |
|
|
|
|
|
|
|
| |
The mount and umount commands will have a bunch of symlinks, so if the
command we are installing is a symlink, preserve it.
Only use -s on install(1) if the command we are installing is an ELF
executable. Since there's a check for that, use it to fire off
get_dso_deps() for only ELF executables.
|
| | |
|
| |
|
|
|
| |
This at least gets to the partitioning screen; didn't finish an install, but it's
closer
|
| |
|
|
|
|
|
| |
This reverts commit 2734cf8a8077f7108bb65b3724af075825d6eb44.
Doing this means that we can't see all of the UI on some netbooks with higher
DPI screens
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
When there are multiple (pre-existing) logical partitions and a logical
partition other then the last partition gets deleted, the number
of the partitions after it changes. Causing getDeviceByName, with the
name based on the new partition number (gotten through parted) to fail.
Specifically this causes a "'NoneType' object has no attribute 'path'"
exception in populate() in partition_gui.py . This patch fixes this by
updating all partitions names in doPartitioning.
Note that for this patch to help the following pyparted bug must also
be fixed: https://fedorahosted.org/pyparted/ticket/17
|
| |
|
|
| |
Missed this one in my initial cciss prep patch.
|
| | |
|
| |
|
|
|
|
|
|
|
|
| |
* We get RAID1 from ks parser, mdadm accepts only lowercase level
specifiers of this form.
* We get int device number (e.g. 0) from ks option --device (e.g. md0).
Though this might require rather changes in pykickstart, especially
when mdadm 3.0 allows creating of arbitrary device names in /dev/md.
* mdadm --create --level option is changed to comply strictly to
documentation (e.g. "--level=raid0", though "--level raid0" works too)
|
| |
|
|
|
| |
We don't really use debug mode for this anymore and it causes problems
in cases like running from the livecd (#493943).
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Our C-code isys EDD BIOS disk order detection is broken in 2 ways atm:
1) It tries to open "sda" instead of "/dev/sda"
2) It tries to open floppies and when it fails, it aborts further device
scanning
Besides that it didn't work with dmraid setups due to the following issues:
3) The C-code will not identify a device by the mbr signature from the
EDD BIOS info, if multiple devices match the signature, however
if we have a fakeraid mirror for example we will have 3 matching devices
both the 2 raw disks and the device mapper device.
4) The python code was using /dev/mapper/longName names where as the
C-code uses /dev/dm-#
5) The python code was calling the C-code (which builds a cache) before it
setup the dmraid sets, so those were not known to the C-code
|
| |
|
|
|
|
|
|
| |
If there was no extended partition originally and we create one, we create
the device for it, but if it changes after that (manual partitioning), we
do not update the created device with the new geometry, leading to a
backtrace with an overlapping partition error when the partition gets
written to the disk.
|
| | |
|
| |
|
|
| |
This is good for when we go back and have to reset the devicetree.
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
| |
We shouldn't be mounting any filesystems in DeviceTree.populate. The
only way we should be able to get an FSError is doing the minSize
calculation, which should not raise exceptions in the first place. If
we do have a way to get filesystem errors, let's find out what they are
and fix them instead of masking them.
|
| |
|
|
|
| |
These should either get a min size or return a reasonable default.
Crashing the installer for something like this is not reasonable.
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
| |
Instead of loading filesystem modules all at once in loader, load them as
needed from the filesystem format __init__ methods. The intention here is to
remove a lot of the special code from loader and avoid kernel errors in
modules that the user never even wants to have involved.
|
| | |
|
| |
|
|
|
|
|
| |
Store mapping from pv names used in ks (e.g. pv.1, raid.1) to existing
partitions in ks handler and use it when processing ks commands referring
to the names.
Also fix some attribute references.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
| |
This should mean that the serial device and options once again get preserved
in the installed system's bootloader config.
|
| | |
|
| |
|
|
|
|
|
|
| |
We call allocatePartitions() multiple times (from growPartitions for example),
and at the top we unalloc partitions, when reallocating them, the
numbering could change and allocatePartitions() may fail half way leaving
us with identical named devices. Reset the PartitionDevice name to req#
when unallocating the partition to fix this.
|
| |
|
|
|
|
|
|
|
|
|
|
| |
parted always renumbers partitions when a partition gets removed, so
if we have p5 and p6, and then p5 gets removed then p6 becomes p5, if
we then check for p6 using getPartitionByPath() it will show up as
non existing even though it does exist, but it has been renamed to p5.
This patch fixes this by simply setting the PartitionDevice's pedPartition
to None when it was removed from its disk, as the PartitionDevice is of no
use to use after that anyways (and we already check that we do not try
to remove None from the disk).
|
| | |
|
| |
|
|
|
|
| |
This hopefully fixes cciss (will test once an anaconda with this in
has hit rawhide, so I can use rhts to test). This is the second version of
this patch, this time it also does the / -> ! translation for sysfs names.
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
This provides a handy way to obtain sizes of devices.
|
