1 |
On Monday, September 18, 2017, Peter Humphrey <peter@××××××××××××.uk> wrote: |
2 |
> |
3 |
> On Monday, 18 September 2017 05:17:34 BST R0b0t1 wrote: |
4 |
> > On Sun, Sep 17, 2017 at 9:12 AM, Peter Humphrey <peter@××××××××××××.uk> |
5 |
> wrote: |
6 |
> > > On Thursday, 14 September 2017 19:51:37 BST R0b0t1 wrote: |
7 |
> > >> On Thu, Sep 14, 2017 at 3:20 AM, Peter Humphrey < |
8 |
peter@××××××××××××.uk> |
9 |
> > > |
10 |
> > > wrote: |
11 |
> > >> > On Thursday, 14 September 2017 05:09:14 BST R0b0t1 wrote: |
12 |
> > >> >> The trickiest part is still the same - going from GRUB or, now, |
13 |
your |
14 |
> > >> >> EFI shell, to Window's bootloader. See here: |
15 |
> > >> >> https://wiki.archlinux.org/index.php/GRUB#Chainloading_ |
16 |
Windows.2FLin |
17 |
> > >> >> ux_ |
18 |
> > >> >> ins talled_in_UEFI_mode. |
19 |
> > >> > |
20 |
> > >> > That advice, though helpful, is about Grub, which isn't installed |
21 |
on |
22 |
> > >> > this box. I did try at first to get it to work here, but failed, |
23 |
so I |
24 |
> > >> > removed it and went for bootctl. It's a fiddle to keep up to date |
25 |
> > >> > with |
26 |
> > >> > kernel upgrades, but at least it works. |
27 |
> > >> |
28 |
> > >> In that case it seems like systemd-boot will check for the Windows |
29 |
> > >> loader and add it to its menu automatically |
30 |
> > >> (https://wiki.archlinux.org/index.php/systemd-boot#Adding_ |
31 |
boot_entries) |
32 |
> > >> . |
33 |
> > >> As above, you may need to reinstall it if the Windows bootloader |
34 |
> > >> installs itself on top of systemd-boot. |
35 |
> > >> |
36 |
> > >> I originally thought you were just booting an EFI stub kernel, in |
37 |
> > >> which case you would have needed some kind of boot manager. |
38 |
> > > |
39 |
> > > I have three questions now: |
40 |
> > > |
41 |
> > > 1. Will Windows 10 install itself in the unpartitioned space? |
42 |
I've |
43 |
> > > attached a screen shot of gparted to show the current layout. |
44 |
> > |
45 |
> > Yes. It will split the free space into a number of partitions if you |
46 |
> > give the installer no further instruction besides selecting the |
47 |
> > unallocated area. |
48 |
> |
49 |
> That's what I was hoping to hear - thanks. |
50 |
> |
51 |
> > To force Windows to use one partition delete the ones it creates |
52 |
> > automatically. You will need to select "custom" or "advanced" in every |
53 |
> > place it is offered as an option. |
54 |
> > |
55 |
> > > 2. What will happen to the UEFI kernel entries in /dev/nvme0n1p1? |
56 |
> > |
57 |
> > When people say "entries" they are usually referring to settings in |
58 |
> > the nonvolatile memory used by a motherboard's EFI firmware. An entry |
59 |
> > associates with an ID a path, priority, and name which is used to |
60 |
> > start the corresponding EFI executable. |
61 |
> |
62 |
> I mean the things that "bootctl status" displays. I've already disabled |
63 |
the |
64 |
> unwanted ones in the UEFI BIOS's list of bootable kernels, but bootctl |
65 |
still |
66 |
> shows them and won't remove them. |
67 |
> |
68 |
|
69 |
Having checked bootctl's documentation it should be changing EFI variables |
70 |
(it may manage kernels also, I am not entirely sure). Are you sure this |
71 |
isn't related to the bug Mick mentioned? If it is then I am unsure why |
72 |
efibootmgr works. |
73 |
|
74 |
Now it's fixed (by using something else) and I can't expect you to care, |
75 |
but I am left perplexed. |
76 |
|
77 |
> > The actual kernels on /dev/nvme0np1 will remain there because Windows |
78 |
> > won't touch that partition unless you tell it to. |
79 |
> > |
80 |
> > > 3. Those entries include some left over from experimenting with |
81 |
> > > other distros. How can I manage the entries and purge the ones I don't |
82 |
> > > need? "Bootctl remove" ignores them. |
83 |
> > |
84 |
> > If you are referring to the kernels left in your /boot then simply |
85 |
> > delete them. "Bootctl remove" and other EFI boot managers I have seen |
86 |
> > refuse to touch your disk. They operate on the EFI configuration |
87 |
> > memory. |
88 |
> > |
89 |
> > > Thanks everyone for your help so far. |
90 |
> > > |
91 |
> > > I don't want to install into a VM, because my main reason for |
92 |
installing |
93 |
> > > Win10 is to be able to run an occasional firmware update program, none |
94 |
> > > of |
95 |
> > > which, it seems, run on Linux. Of course, it should also help me get |
96 |
up |
97 |
> > > to speed with the M$ world. |
98 |
> > |
99 |
> > If you pass an entire hard disk to the VM you can then take it out and |
100 |
> > put it in another computer and boot it (or boot it in the same |
101 |
> > computer sans hypervisor). |
102 |
> |
103 |
> Maybe that's a use for a couple of spare SSDs I have here. |
104 |
> |
105 |
> > With Linux you can pass partitions in individually and use what the |
106 |
> > guest thinks is a raw character device as a disk, so that if you |
107 |
> > wanted to boot that installation from outside of the hypervisor you |
108 |
> > could. This might not be possible with Windows. |
109 |
> > |
110 |
> > If you install into a VM you can pass almost everything to the VM |
111 |
> > directly. I suppose the only thing that may not work extremely well |
112 |
> > would be motherboard firmware updates, but if you look QEMU has |
113 |
> > options to pass almost everything in a computer to a VM. Admittedly |
114 |
> > this isn't a very plug-and-play solution. |
115 |
> > |
116 |
> > Aside from firmware updates (realize though that almost everything - |
117 |
> > barring some low level interfaces like I2C - can be passed to a VM) I |
118 |
> > would invite you to use Windows only in a VM. I find it easier to get |
119 |
> > work done in this way while using Windows programs. Xfreerdp is a good |
120 |
> > way to interact with a Windows guest and can provide better desktop |
121 |
> > integration than QEMU or libvirtd. |
122 |
> |
123 |
> I use VirtualBox here, mostly because some BOINC projects require it. |
124 |
> |
125 |
|
126 |
Fair enough. |