#2728 PR merged: Always include loaded kernel modules¶
Labels: bug, fixed / solved / done
jsmeix opened issue at 2021-12-15 12:29:¶
-
Type: Bug Fix
-
Impact: Normal
-
Reference to related issue (URL):
https://github.com/rear/rear/issues/2727 -
How was this pull request tested?
Seems to work OK for me - but I can only do a few generic tests.
In particular withMODULES=( 'loaded_modules' 'wireguard' )
I get exactly only the loaded kernel modules but no 'wireguard'. -
Brief description of the changes in this pull request:
In build/GNU/Linux/400_copy_modules.sh also
for MODULES=() or MODULES=( 'moduleX' 'moduleY' )
include the currently loaded kernel modules.
jsmeix commented at 2021-12-15 12:32:¶
@DEvil0000
could you test it?
I would much appreciate it if you could.
With the change here
MODULES=( 'moduleX' 'moduleY' )
should automatically also include all loaded kernel modules
(plus some more - see the 400_copy_modules.sh code here).
DEvil0000 commented at 2021-12-15 16:45:¶
MODULES=( 'loaded_modules' 'wireguard' ) caused no modules at all got
packed in the ISO. Is this intended?
edit:
MODULES=( 'wireguard' ) did pack/include some loaded modules and
wireguard. It somehow did not include module uas which was also
loaded.
yes, I am aware that the recovery system does not need to load the same
modules - but some are missing in the image.
lsmod before rear mkrescue:
root@vl1bbls1a:~# lsmod
Module Size Used by
vxlan 69632 0
ip6_udp_tunnel 16384 1 vxlan
udp_tunnel 16384 1 vxlan
uio 20480 0
pf_ring 724992 0
cfg80211 708608 0
dm_multipath 32768 0
scsi_dh_rdac 16384 0
scsi_dh_emc 16384 0
scsi_dh_alua 20480 0
amd64_edac_mod 36864 0
edac_mce_amd 32768 1 amd64_edac_mod
kvm_amd 98304 0
kvm 663552 1 kvm_amd
pcengines_apuv2 16384 0
gpio_keys_polled 16384 0
input_polldev 20480 1 gpio_keys_polled
gpio_amd_fch 16384 3
k10temp 16384 0
fam15h_power 16384 0
ccp 86016 11 kvm_amd
mac_hid 16384 0
leds_gpio 16384 0
sch_fq_codel 20480 14
sctp 344064 2
gpio_nct5104d 16384 1
8021q 32768 0
garp 16384 1 8021q
mrp 20480 1 8021q
stp 16384 1 garp
llc 16384 2 stp,garp
bonding 167936 0
drm 491520 0
ip_tables 32768 0
x_tables 40960 1 ip_tables
autofs4 45056 2
btrfs 1261568 0
zstd_compress 167936 1 btrfs
raid10 61440 0
raid456 155648 0
async_raid6_recov 24576 1 raid456
async_memcpy 20480 2 raid456,async_raid6_recov
async_pq 24576 2 raid456,async_raid6_recov
async_xor 20480 3 async_pq,raid456,async_raid6_recov
async_tx 20480 5 async_pq,async_memcpy,async_xor,raid456,async_raid6_recov
xor 24576 2 async_xor,btrfs
raid6_pq 114688 4 async_pq,btrfs,raid456,async_raid6_recov
libcrc32c 16384 3 btrfs,raid456,sctp
raid1 45056 0
raid0 24576 0
multipath 20480 0
linear 20480 0
crct10dif_pclmul 16384 1
crc32_pclmul 16384 0
ghash_clmulni_intel 16384 0
aesni_intel 372736 0
crypto_simd 16384 1 aesni_intel
uas 28672 0
sdhci_pci 53248 0
cryptd 24576 2 crypto_simd,ghash_clmulni_intel
glue_helper 16384 1 aesni_intel
cqhci 28672 1 sdhci_pci
ahci 40960 2
libahci 32768 1 ahci
sdhci 65536 1 sdhci_pci
r8169 90112 0
igb 221184 0
usb_storage 77824 1 uas
i2c_piix4 28672 0
realtek 24576 1
dca 16384 1 igb
i2c_algo_bit 16384 1 igb
gpio_keys 20480 0
lsmod when rear ISO is booted:
Module Size Used by
nvram 16384 0
sctp 344064 2
libcrc32c 16384 1 sctp
gpio_nct5104d 16384 1
8021q 32768 0
garp 16384 1 8021q
mrp 20480 1 8021q
stp 16384 1 garp
llc 16384 2 stp,garp
bonding 167936 0
amd64_edac_mod 36864 0
edac_mce_amd 32768 1 amd64_edac_mod
kvm_amd 98304 0
kvm 663552 1 kvm_amd
crct10dif_pclmul 16384 1
crc32_pclmul 16384 0
ghash_clmulni_intel 16384 0
pcengines_apuv2 16384 0
gpio_keys_polled 16384 0
aesni_intel 372736 0
input_polldev 20480 1 gpio_keys_polled
gpio_amd_fch 16384 3
glue_helper 16384 1 aesni_intel
crypto_simd 16384 1 aesni_intel
cryptd 24576 2 crypto_simd,ghash_clmulni_intel
sdhci_pci 53248 0
cqhci 28672 1 sdhci_pci
usb_storage 77824 0
ahci 40960 0
r8169 90112 0
k10temp 16384 0
igb 221184 0
i2c_piix4 28672 0
libahci 32768 1 ahci
fam15h_power 16384 0
realtek 24576 0
sdhci 65536 1 sdhci_pci
i2c_algo_bit 16384 1 igb
ccp 86016 1 kvm_amd
dca 16384 1 igb
mac_hid 16384 0
leds_gpio 16384 0
gpio_keys 20480 0
ip_tables 32768 0
x_tables 40960 1 ip_tables
autofs4 45056 0
I see a modprobe error on boot and modprobe uas does not find the named module.
jsmeix commented at 2021-12-16 09:04:¶
This specific modules including code in 400_copy_modules.sh
was and still is an endless nightmare.
The only thing that works simple and straightforward is the
MODULES=( 'all_modules' ) case because that does not
need the horrible modinfo_filename() function
which is the actual nightmare because - as far as I know - there is
no simple and straightforward way how to reliably get the right
kernel module file with path (i.e. what we need to copy)
from a kernel module name (i.e. what the user specified or what lsmod
shows)
Currently we have the issue
https://github.com/rear/rear/issues/2677
Perhaps you are also hit by it?
DEvil0000 commented at 2021-12-16 13:07:¶
I think the issue with uas is with line
176.
grep of uas and similar short strings are likely to match somewhere
even if this is not the module we are looking for. In case of uas it
matcheskernel/fs/squashfs/squashfs.ko. I suggest changing the grep to
/uas.ko so prepending / and appending .ko.
I am not sure if this is a wise thing to do anyway - read second part
about #2677.
I think this issue is related to #2677 but not actually the same bug.
modinfo should be fine to search modules (if depmod -a or similar was
run correctly before so lookup is fine?). The lookup order for module
pathes is defined in a way that build in modules can get 'updated' by
packages providing a newer version of the module. So even if it is a
built-in one you should first check for a newer version actually used
(modinfo maybe)
edit:
please note that lsmod will not show built-in modules normally (loaded
or not does not matter)
hpannenb commented at 2021-12-16 19:13:¶
I agree with @DEvil0000 : The mentioned grep is a too 'lazy' and would
lead to the mentioned false positives.
I propose to change the line to something like
grep -q "\/$( echo $module | tr '_-' '..' )\." ...
to search with a prefix / and an extension of a single dot to be open
for any module name's extension that might come.
Edit:
The only issue I see is if $module is empty code would continue since
there is no entry found in the modules.builtin.
jsmeix commented at 2021-12-22 13:00:¶
@DEvil0000 @hpannenb
I will work on this one (as time permits) in January.
I wish you a merry Christmas and a happy New Year!
github-actions commented at 2022-02-21 02:09:¶
Stale pull request message
github-actions commented at 2022-04-23 02:49:¶
Stale pull request message
jsmeix commented at 2022-05-16 10:24:¶
With
https://github.com/rear/rear/pull/2728/commits/048db085c1cb0a81186d9675b4126d92223483ab
I have now
grep -q "/$( echo $module | tr '_-' '..' )\." /lib/modules/$KERNEL_VERSION/modules.builtin && continue
There is no need to quote '/' with a preceding backslash
because '/' is no grep meta-character as far as I see
(I did not find a complete list of grep meta-characters)
in contrast to '.' which is a grep meta-character.
There is no need to care about if $module is empty
because the line just before
modinfo $module 1>/dev/null || continue
does that:
# modinfo
modinfo: ERROR: missing module or filename.
jsmeix commented at 2022-05-16 10:33:¶
@rear/contributors
please have a look here if you see obviously wrong changes.
When there are no objections I would like to merge it
on Wednesday (May 18) afternoon.
DEvil0000 commented at 2022-05-17 14:22:¶
Changeset looks fine, I however currently have no time to test it.
[Export of Github issue for rear/rear.]